diff --git "a/model.safetensors.index.json" "b/model.safetensors.index.json"
new file mode 100644--- /dev/null
+++ "b/model.safetensors.index.json"
@@ -0,0 +1,6212 @@
+{
+    "metadata": {
+        "total_size": 74910068736
+    },
+    "weight_map": {
+        "lm_head.biases": "model-00014-of-00014.safetensors",
+        "lm_head.scales": "model-00014-of-00014.safetensors",
+        "lm_head.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.0.v1.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.0.v1.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.0.v1.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.0.w1.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.0.w1.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.0.w1.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.0.w2.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.0.w2.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.0.w2.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.1.v1.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.1.v1.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.1.v1.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.1.w1.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.1.w1.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.1.w1.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.1.w2.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.1.w2.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.1.w2.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.10.v1.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.10.v1.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.10.v1.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.10.w1.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.10.w1.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.10.w1.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.10.w2.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.10.w2.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.10.w2.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.11.v1.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.11.v1.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.11.v1.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.11.w1.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.11.w1.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.11.w1.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.11.w2.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.11.w2.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.11.w2.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.12.v1.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.12.v1.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.12.v1.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.12.w1.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.12.w1.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.12.w1.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.12.w2.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.12.w2.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.12.w2.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.13.v1.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.13.v1.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.13.v1.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.13.w1.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.13.w1.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.13.w1.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.13.w2.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.13.w2.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.13.w2.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.14.v1.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.14.v1.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.14.v1.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.14.w1.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.14.w1.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.14.w1.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.14.w2.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.14.w2.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.14.w2.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.15.v1.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.15.v1.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.15.v1.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.15.w1.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.15.w1.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.15.w1.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.15.w2.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.15.w2.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.15.w2.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.2.v1.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.2.v1.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.2.v1.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.2.w1.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.2.w1.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.2.w1.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.2.w2.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.2.w2.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.2.w2.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.3.v1.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.3.v1.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.3.v1.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.3.w1.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.3.w1.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.3.w1.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.3.w2.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.3.w2.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.3.w2.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.4.v1.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.4.v1.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.4.v1.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.4.w1.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.4.w1.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.4.w1.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.4.w2.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.4.w2.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.4.w2.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.5.v1.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.5.v1.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.5.v1.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.5.w1.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.5.w1.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.5.w1.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.5.w2.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.5.w2.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.5.w2.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.6.v1.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.6.v1.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.6.v1.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.6.w1.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.6.w1.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.6.w1.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.6.w2.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.6.w2.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.6.w2.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.7.v1.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.7.v1.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.7.v1.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.7.w1.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.7.w1.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.7.w1.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.7.w2.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.7.w2.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.7.w2.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.8.v1.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.8.v1.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.8.v1.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.8.w1.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.8.w1.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.8.w1.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.8.w2.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.8.w2.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.8.w2.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.9.v1.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.9.v1.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.9.v1.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.9.w1.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.9.w1.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.9.w1.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.9.w2.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.9.w2.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.experts.9.w2.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.router.layer.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.router.layer.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.ffn.router.layer.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.norm_attn_norm.attn.Wqkv.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.norm_attn_norm.attn.Wqkv.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.norm_attn_norm.attn.Wqkv.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.norm_attn_norm.attn.out_proj.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.norm_attn_norm.attn.out_proj.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.norm_attn_norm.attn.out_proj.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.norm_attn_norm.norm_1.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.0.norm_attn_norm.norm_2.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.0.v1.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.0.v1.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.0.v1.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.0.w1.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.0.w1.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.0.w1.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.0.w2.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.0.w2.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.0.w2.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.1.v1.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.1.v1.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.1.v1.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.1.w1.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.1.w1.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.1.w1.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.1.w2.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.1.w2.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.1.w2.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.10.v1.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.10.v1.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.10.v1.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.10.w1.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.10.w1.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.10.w1.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.10.w2.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.10.w2.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.10.w2.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.11.v1.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.11.v1.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.11.v1.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.11.w1.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.11.w1.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.11.w1.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.11.w2.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.11.w2.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.11.w2.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.12.v1.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.12.v1.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.12.v1.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.12.w1.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.12.w1.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.12.w1.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.12.w2.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.12.w2.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.12.w2.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.13.v1.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.13.v1.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.13.v1.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.13.w1.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.13.w1.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.13.w1.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.13.w2.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.13.w2.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.13.w2.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.14.v1.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.14.v1.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.14.v1.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.14.w1.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.14.w1.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.14.w1.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.14.w2.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.14.w2.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.14.w2.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.15.v1.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.15.v1.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.15.v1.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.15.w1.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.15.w1.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.15.w1.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.15.w2.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.15.w2.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.15.w2.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.2.v1.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.2.v1.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.2.v1.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.2.w1.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.2.w1.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.2.w1.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.2.w2.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.2.w2.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.2.w2.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.3.v1.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.3.v1.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.3.v1.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.3.w1.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.3.w1.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.3.w1.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.3.w2.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.3.w2.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.3.w2.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.4.v1.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.4.v1.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.4.v1.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.4.w1.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.4.w1.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.4.w1.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.4.w2.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.4.w2.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.4.w2.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.5.v1.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.5.v1.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.5.v1.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.5.w1.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.5.w1.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.5.w1.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.5.w2.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.5.w2.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.5.w2.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.6.v1.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.6.v1.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.6.v1.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.6.w1.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.6.w1.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.6.w1.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.6.w2.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.6.w2.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.6.w2.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.7.v1.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.7.v1.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.7.v1.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.7.w1.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.7.w1.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.7.w1.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.7.w2.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.7.w2.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.7.w2.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.8.v1.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.8.v1.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.8.v1.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.8.w1.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.8.w1.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.8.w1.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.8.w2.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.8.w2.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.8.w2.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.9.v1.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.9.v1.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.9.v1.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.9.w1.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.9.w1.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.9.w1.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.9.w2.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.9.w2.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.experts.9.w2.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.router.layer.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.router.layer.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.ffn.router.layer.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.norm_attn_norm.attn.Wqkv.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.norm_attn_norm.attn.Wqkv.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.norm_attn_norm.attn.Wqkv.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.norm_attn_norm.attn.out_proj.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.norm_attn_norm.attn.out_proj.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.norm_attn_norm.attn.out_proj.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.norm_attn_norm.norm_1.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.1.norm_attn_norm.norm_2.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.0.v1.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.0.v1.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.0.v1.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.0.w1.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.0.w1.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.0.w1.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.0.w2.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.0.w2.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.0.w2.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.1.v1.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.1.v1.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.1.v1.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.1.w1.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.1.w1.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.1.w1.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.1.w2.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.1.w2.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.1.w2.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.10.v1.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.10.v1.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.10.v1.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.10.w1.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.10.w1.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.10.w1.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.10.w2.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.10.w2.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.10.w2.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.11.v1.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.11.v1.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.11.v1.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.11.w1.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.11.w1.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.11.w1.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.11.w2.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.11.w2.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.11.w2.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.12.v1.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.12.v1.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.12.v1.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.12.w1.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.12.w1.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.12.w1.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.12.w2.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.12.w2.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.12.w2.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.13.v1.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.13.v1.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.13.v1.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.13.w1.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.13.w1.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.13.w1.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.13.w2.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.13.w2.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.13.w2.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.14.v1.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.14.v1.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.14.v1.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.14.w1.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.14.w1.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.14.w1.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.14.w2.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.14.w2.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.14.w2.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.15.v1.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.15.v1.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.15.v1.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.15.w1.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.15.w1.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.15.w1.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.15.w2.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.15.w2.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.15.w2.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.2.v1.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.2.v1.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.2.v1.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.2.w1.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.2.w1.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.2.w1.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.2.w2.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.2.w2.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.2.w2.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.3.v1.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.3.v1.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.3.v1.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.3.w1.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.3.w1.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.3.w1.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.3.w2.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.3.w2.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.3.w2.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.4.v1.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.4.v1.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.4.v1.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.4.w1.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.4.w1.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.4.w1.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.4.w2.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.4.w2.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.4.w2.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.5.v1.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.5.v1.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.5.v1.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.5.w1.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.5.w1.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.5.w1.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.5.w2.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.5.w2.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.5.w2.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.6.v1.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.6.v1.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.6.v1.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.6.w1.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.6.w1.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.6.w1.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.6.w2.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.6.w2.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.6.w2.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.7.v1.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.7.v1.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.7.v1.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.7.w1.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.7.w1.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.7.w1.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.7.w2.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.7.w2.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.7.w2.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.8.v1.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.8.v1.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.8.v1.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.8.w1.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.8.w1.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.8.w1.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.8.w2.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.8.w2.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.8.w2.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.9.v1.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.9.v1.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.9.v1.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.9.w1.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.9.w1.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.9.w1.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.9.w2.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.9.w2.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.experts.9.w2.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.router.layer.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.router.layer.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.ffn.router.layer.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.norm_attn_norm.attn.Wqkv.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.norm_attn_norm.attn.Wqkv.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.norm_attn_norm.attn.Wqkv.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.norm_attn_norm.attn.out_proj.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.norm_attn_norm.attn.out_proj.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.norm_attn_norm.attn.out_proj.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.norm_attn_norm.norm_1.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.10.norm_attn_norm.norm_2.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.0.v1.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.0.v1.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.0.v1.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.0.w1.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.0.w1.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.0.w1.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.0.w2.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.0.w2.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.0.w2.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.1.v1.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.1.v1.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.1.v1.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.1.w1.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.1.w1.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.1.w1.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.1.w2.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.1.w2.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.1.w2.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.10.v1.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.10.v1.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.10.v1.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.10.w1.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.10.w1.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.10.w1.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.10.w2.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.10.w2.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.10.w2.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.11.v1.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.11.v1.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.11.v1.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.11.w1.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.11.w1.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.11.w1.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.11.w2.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.11.w2.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.11.w2.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.12.v1.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.12.v1.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.12.v1.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.12.w1.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.12.w1.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.12.w1.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.12.w2.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.12.w2.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.12.w2.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.13.v1.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.13.v1.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.13.v1.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.13.w1.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.13.w1.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.13.w1.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.13.w2.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.13.w2.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.13.w2.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.14.v1.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.14.v1.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.14.v1.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.14.w1.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.14.w1.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.14.w1.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.14.w2.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.14.w2.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.14.w2.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.15.v1.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.15.v1.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.15.v1.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.15.w1.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.15.w1.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.15.w1.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.15.w2.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.15.w2.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.15.w2.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.2.v1.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.2.v1.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.2.v1.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.2.w1.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.2.w1.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.2.w1.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.2.w2.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.2.w2.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.2.w2.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.3.v1.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.3.v1.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.3.v1.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.3.w1.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.3.w1.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.3.w1.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.3.w2.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.3.w2.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.3.w2.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.4.v1.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.4.v1.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.4.v1.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.4.w1.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.4.w1.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.4.w1.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.4.w2.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.4.w2.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.4.w2.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.5.v1.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.5.v1.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.5.v1.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.5.w1.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.5.w1.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.5.w1.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.5.w2.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.5.w2.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.5.w2.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.6.v1.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.6.v1.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.6.v1.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.6.w1.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.6.w1.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.6.w1.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.6.w2.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.6.w2.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.6.w2.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.7.v1.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.7.v1.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.7.v1.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.7.w1.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.7.w1.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.7.w1.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.7.w2.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.7.w2.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.7.w2.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.8.v1.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.8.v1.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.8.v1.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.8.w1.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.8.w1.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.8.w1.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.8.w2.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.8.w2.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.8.w2.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.9.v1.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.9.v1.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.9.v1.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.9.w1.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.9.w1.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.9.w1.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.9.w2.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.9.w2.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.experts.9.w2.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.ffn.router.layer.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.11.ffn.router.layer.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.11.ffn.router.layer.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.11.norm_attn_norm.attn.Wqkv.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.norm_attn_norm.attn.Wqkv.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.norm_attn_norm.attn.Wqkv.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.norm_attn_norm.attn.out_proj.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.norm_attn_norm.attn.out_proj.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.norm_attn_norm.attn.out_proj.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.norm_attn_norm.norm_1.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.11.norm_attn_norm.norm_2.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.0.v1.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.0.v1.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.0.v1.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.0.w1.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.0.w1.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.0.w1.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.0.w2.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.0.w2.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.0.w2.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.1.v1.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.1.v1.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.1.v1.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.1.w1.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.1.w1.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.1.w1.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.1.w2.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.1.w2.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.1.w2.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.10.v1.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.10.v1.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.10.v1.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.10.w1.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.10.w1.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.10.w1.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.10.w2.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.10.w2.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.10.w2.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.11.v1.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.11.v1.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.11.v1.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.11.w1.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.11.w1.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.11.w1.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.11.w2.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.11.w2.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.11.w2.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.12.v1.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.12.v1.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.12.v1.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.12.w1.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.12.w1.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.12.w1.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.12.w2.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.12.w2.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.12.w2.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.13.v1.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.13.v1.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.13.v1.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.13.w1.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.13.w1.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.13.w1.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.13.w2.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.13.w2.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.13.w2.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.14.v1.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.14.v1.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.14.v1.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.14.w1.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.14.w1.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.14.w1.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.14.w2.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.14.w2.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.14.w2.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.15.v1.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.15.v1.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.15.v1.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.15.w1.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.15.w1.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.15.w1.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.15.w2.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.15.w2.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.15.w2.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.2.v1.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.2.v1.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.2.v1.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.2.w1.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.2.w1.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.2.w1.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.2.w2.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.2.w2.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.2.w2.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.3.v1.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.3.v1.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.3.v1.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.3.w1.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.3.w1.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.3.w1.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.3.w2.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.3.w2.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.3.w2.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.4.v1.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.4.v1.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.4.v1.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.4.w1.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.4.w1.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.4.w1.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.4.w2.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.4.w2.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.4.w2.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.5.v1.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.5.v1.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.5.v1.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.5.w1.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.5.w1.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.5.w1.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.5.w2.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.5.w2.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.5.w2.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.6.v1.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.6.v1.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.6.v1.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.6.w1.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.6.w1.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.6.w1.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.6.w2.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.6.w2.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.6.w2.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.7.v1.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.7.v1.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.7.v1.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.7.w1.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.7.w1.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.7.w1.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.7.w2.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.7.w2.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.7.w2.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.8.v1.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.8.v1.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.8.v1.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.8.w1.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.8.w1.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.8.w1.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.8.w2.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.8.w2.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.8.w2.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.9.v1.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.9.v1.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.9.v1.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.9.w1.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.9.w1.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.9.w1.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.9.w2.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.9.w2.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.experts.9.w2.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.router.layer.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.router.layer.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.ffn.router.layer.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.norm_attn_norm.attn.Wqkv.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.norm_attn_norm.attn.Wqkv.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.norm_attn_norm.attn.Wqkv.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.norm_attn_norm.attn.out_proj.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.norm_attn_norm.attn.out_proj.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.norm_attn_norm.attn.out_proj.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.norm_attn_norm.norm_1.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.12.norm_attn_norm.norm_2.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.0.v1.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.0.v1.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.0.v1.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.0.w1.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.0.w1.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.0.w1.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.0.w2.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.0.w2.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.0.w2.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.1.v1.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.1.v1.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.1.v1.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.1.w1.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.1.w1.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.1.w1.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.1.w2.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.1.w2.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.1.w2.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.10.v1.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.10.v1.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.10.v1.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.10.w1.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.10.w1.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.10.w1.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.10.w2.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.10.w2.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.10.w2.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.11.v1.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.11.v1.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.11.v1.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.11.w1.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.11.w1.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.11.w1.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.11.w2.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.11.w2.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.11.w2.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.12.v1.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.12.v1.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.12.v1.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.12.w1.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.12.w1.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.12.w1.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.12.w2.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.12.w2.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.12.w2.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.13.v1.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.13.v1.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.13.v1.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.13.w1.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.13.w1.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.13.w1.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.13.w2.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.13.w2.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.13.w2.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.14.v1.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.14.v1.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.14.v1.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.14.w1.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.14.w1.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.14.w1.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.14.w2.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.14.w2.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.14.w2.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.15.v1.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.15.v1.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.15.v1.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.15.w1.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.15.w1.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.15.w1.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.15.w2.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.15.w2.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.15.w2.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.2.v1.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.2.v1.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.2.v1.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.2.w1.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.2.w1.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.2.w1.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.2.w2.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.2.w2.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.2.w2.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.3.v1.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.3.v1.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.3.v1.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.3.w1.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.3.w1.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.3.w1.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.3.w2.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.3.w2.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.3.w2.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.4.v1.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.4.v1.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.4.v1.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.4.w1.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.4.w1.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.4.w1.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.4.w2.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.4.w2.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.4.w2.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.5.v1.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.5.v1.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.5.v1.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.5.w1.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.5.w1.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.5.w1.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.5.w2.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.5.w2.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.5.w2.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.6.v1.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.6.v1.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.6.v1.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.6.w1.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.6.w1.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.6.w1.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.6.w2.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.6.w2.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.6.w2.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.7.v1.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.7.v1.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.7.v1.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.7.w1.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.7.w1.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.7.w1.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.7.w2.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.7.w2.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.7.w2.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.8.v1.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.8.v1.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.8.v1.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.8.w1.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.8.w1.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.8.w1.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.8.w2.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.8.w2.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.8.w2.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.9.v1.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.9.v1.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.9.v1.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.9.w1.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.9.w1.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.9.w1.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.9.w2.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.9.w2.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.experts.9.w2.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.router.layer.biases": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.router.layer.scales": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.ffn.router.layer.weight": "model-00005-of-00014.safetensors",
+        "transformer.blocks.13.norm_attn_norm.attn.Wqkv.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.13.norm_attn_norm.attn.Wqkv.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.13.norm_attn_norm.attn.Wqkv.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.13.norm_attn_norm.attn.out_proj.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.13.norm_attn_norm.attn.out_proj.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.13.norm_attn_norm.attn.out_proj.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.13.norm_attn_norm.norm_1.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.13.norm_attn_norm.norm_2.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.0.v1.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.0.v1.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.0.v1.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.0.w1.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.0.w1.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.0.w1.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.0.w2.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.0.w2.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.0.w2.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.1.v1.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.1.v1.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.1.v1.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.1.w1.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.1.w1.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.1.w1.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.1.w2.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.1.w2.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.1.w2.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.10.v1.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.10.v1.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.10.v1.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.10.w1.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.10.w1.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.10.w1.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.10.w2.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.10.w2.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.10.w2.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.11.v1.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.11.v1.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.11.v1.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.11.w1.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.11.w1.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.11.w1.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.11.w2.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.11.w2.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.11.w2.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.12.v1.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.12.v1.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.12.v1.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.12.w1.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.12.w1.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.12.w1.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.12.w2.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.12.w2.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.12.w2.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.13.v1.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.13.v1.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.13.v1.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.13.w1.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.13.w1.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.13.w1.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.13.w2.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.13.w2.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.13.w2.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.14.v1.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.14.v1.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.14.v1.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.14.w1.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.14.w1.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.14.w1.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.14.w2.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.14.w2.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.14.w2.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.15.v1.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.15.v1.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.15.v1.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.15.w1.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.15.w1.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.15.w1.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.15.w2.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.15.w2.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.15.w2.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.2.v1.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.2.v1.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.2.v1.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.2.w1.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.2.w1.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.2.w1.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.2.w2.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.2.w2.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.2.w2.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.3.v1.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.3.v1.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.3.v1.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.3.w1.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.3.w1.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.3.w1.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.3.w2.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.3.w2.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.3.w2.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.4.v1.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.4.v1.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.4.v1.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.4.w1.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.4.w1.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.4.w1.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.4.w2.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.4.w2.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.4.w2.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.5.v1.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.5.v1.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.5.v1.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.5.w1.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.5.w1.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.5.w1.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.5.w2.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.5.w2.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.5.w2.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.6.v1.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.6.v1.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.6.v1.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.6.w1.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.6.w1.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.6.w1.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.6.w2.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.6.w2.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.6.w2.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.7.v1.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.7.v1.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.7.v1.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.7.w1.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.7.w1.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.7.w1.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.7.w2.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.7.w2.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.7.w2.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.8.v1.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.8.v1.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.8.v1.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.8.w1.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.8.w1.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.8.w1.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.8.w2.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.8.w2.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.8.w2.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.9.v1.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.9.v1.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.9.v1.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.9.w1.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.9.w1.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.9.w1.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.9.w2.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.9.w2.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.experts.9.w2.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.router.layer.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.router.layer.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.ffn.router.layer.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.norm_attn_norm.attn.Wqkv.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.norm_attn_norm.attn.Wqkv.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.norm_attn_norm.attn.Wqkv.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.norm_attn_norm.attn.out_proj.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.norm_attn_norm.attn.out_proj.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.norm_attn_norm.attn.out_proj.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.norm_attn_norm.norm_1.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.14.norm_attn_norm.norm_2.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.0.v1.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.0.v1.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.0.v1.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.0.w1.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.0.w1.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.0.w1.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.0.w2.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.0.w2.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.0.w2.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.1.v1.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.1.v1.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.1.v1.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.1.w1.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.1.w1.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.1.w1.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.1.w2.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.1.w2.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.1.w2.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.10.v1.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.10.v1.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.10.v1.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.10.w1.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.10.w1.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.10.w1.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.10.w2.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.10.w2.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.10.w2.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.11.v1.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.11.v1.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.11.v1.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.11.w1.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.11.w1.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.11.w1.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.11.w2.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.11.w2.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.11.w2.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.12.v1.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.12.v1.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.12.v1.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.12.w1.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.12.w1.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.12.w1.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.12.w2.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.12.w2.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.12.w2.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.13.v1.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.13.v1.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.13.v1.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.13.w1.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.13.w1.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.13.w1.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.13.w2.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.13.w2.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.13.w2.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.14.v1.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.14.v1.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.14.v1.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.14.w1.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.14.w1.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.14.w1.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.14.w2.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.14.w2.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.14.w2.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.15.v1.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.15.v1.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.15.v1.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.15.w1.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.15.w1.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.15.w1.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.15.w2.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.15.w2.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.15.w2.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.2.v1.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.2.v1.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.2.v1.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.2.w1.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.2.w1.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.2.w1.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.2.w2.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.2.w2.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.2.w2.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.3.v1.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.3.v1.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.3.v1.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.3.w1.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.3.w1.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.3.w1.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.3.w2.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.3.w2.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.3.w2.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.4.v1.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.4.v1.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.4.v1.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.4.w1.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.4.w1.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.4.w1.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.4.w2.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.4.w2.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.4.w2.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.5.v1.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.5.v1.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.5.v1.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.5.w1.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.5.w1.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.5.w1.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.5.w2.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.5.w2.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.5.w2.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.6.v1.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.6.v1.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.6.v1.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.6.w1.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.6.w1.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.6.w1.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.6.w2.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.6.w2.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.6.w2.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.7.v1.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.7.v1.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.7.v1.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.7.w1.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.7.w1.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.7.w1.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.7.w2.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.7.w2.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.7.w2.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.8.v1.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.8.v1.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.8.v1.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.8.w1.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.8.w1.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.8.w1.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.8.w2.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.8.w2.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.8.w2.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.9.v1.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.9.v1.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.9.v1.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.9.w1.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.9.w1.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.9.w1.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.9.w2.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.9.w2.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.experts.9.w2.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.router.layer.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.router.layer.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.ffn.router.layer.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.norm_attn_norm.attn.Wqkv.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.norm_attn_norm.attn.Wqkv.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.norm_attn_norm.attn.Wqkv.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.norm_attn_norm.attn.out_proj.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.norm_attn_norm.attn.out_proj.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.norm_attn_norm.attn.out_proj.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.norm_attn_norm.norm_1.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.15.norm_attn_norm.norm_2.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.0.v1.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.0.v1.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.0.v1.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.0.w1.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.0.w1.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.0.w1.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.0.w2.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.0.w2.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.0.w2.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.1.v1.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.1.v1.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.1.v1.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.1.w1.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.1.w1.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.1.w1.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.1.w2.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.1.w2.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.1.w2.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.10.v1.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.10.v1.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.10.v1.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.10.w1.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.10.w1.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.10.w1.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.10.w2.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.10.w2.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.10.w2.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.11.v1.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.11.v1.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.11.v1.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.11.w1.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.11.w1.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.11.w1.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.11.w2.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.11.w2.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.11.w2.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.12.v1.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.12.v1.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.12.v1.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.12.w1.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.12.w1.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.12.w1.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.12.w2.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.12.w2.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.12.w2.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.13.v1.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.13.v1.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.13.v1.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.13.w1.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.13.w1.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.13.w1.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.13.w2.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.13.w2.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.13.w2.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.14.v1.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.14.v1.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.14.v1.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.14.w1.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.14.w1.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.14.w1.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.14.w2.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.14.w2.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.14.w2.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.15.v1.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.15.v1.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.15.v1.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.15.w1.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.15.w1.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.15.w1.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.15.w2.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.15.w2.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.15.w2.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.2.v1.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.2.v1.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.2.v1.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.2.w1.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.2.w1.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.2.w1.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.2.w2.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.2.w2.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.2.w2.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.3.v1.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.3.v1.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.3.v1.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.3.w1.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.3.w1.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.3.w1.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.3.w2.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.3.w2.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.3.w2.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.4.v1.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.4.v1.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.4.v1.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.4.w1.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.4.w1.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.4.w1.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.4.w2.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.4.w2.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.4.w2.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.5.v1.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.5.v1.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.5.v1.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.5.w1.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.5.w1.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.5.w1.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.5.w2.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.5.w2.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.5.w2.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.6.v1.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.6.v1.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.6.v1.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.6.w1.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.6.w1.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.6.w1.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.6.w2.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.6.w2.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.6.w2.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.7.v1.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.7.v1.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.7.v1.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.7.w1.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.7.w1.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.7.w1.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.7.w2.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.7.w2.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.7.w2.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.8.v1.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.8.v1.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.8.v1.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.8.w1.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.8.w1.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.8.w1.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.8.w2.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.8.w2.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.8.w2.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.9.v1.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.9.v1.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.9.v1.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.9.w1.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.9.w1.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.9.w1.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.9.w2.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.9.w2.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.experts.9.w2.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.router.layer.biases": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.router.layer.scales": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.ffn.router.layer.weight": "model-00006-of-00014.safetensors",
+        "transformer.blocks.16.norm_attn_norm.attn.Wqkv.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.16.norm_attn_norm.attn.Wqkv.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.16.norm_attn_norm.attn.Wqkv.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.16.norm_attn_norm.attn.out_proj.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.16.norm_attn_norm.attn.out_proj.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.16.norm_attn_norm.attn.out_proj.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.16.norm_attn_norm.norm_1.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.16.norm_attn_norm.norm_2.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.0.v1.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.0.v1.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.0.v1.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.0.w1.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.0.w1.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.0.w1.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.0.w2.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.0.w2.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.0.w2.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.1.v1.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.1.v1.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.1.v1.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.1.w1.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.1.w1.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.1.w1.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.1.w2.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.1.w2.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.1.w2.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.10.v1.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.10.v1.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.10.v1.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.10.w1.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.10.w1.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.10.w1.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.10.w2.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.10.w2.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.10.w2.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.11.v1.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.11.v1.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.11.v1.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.11.w1.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.11.w1.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.11.w1.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.11.w2.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.11.w2.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.11.w2.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.12.v1.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.12.v1.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.12.v1.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.12.w1.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.12.w1.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.12.w1.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.12.w2.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.12.w2.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.12.w2.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.13.v1.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.13.v1.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.13.v1.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.13.w1.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.13.w1.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.13.w1.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.13.w2.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.13.w2.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.13.w2.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.14.v1.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.14.v1.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.14.v1.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.14.w1.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.14.w1.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.14.w1.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.14.w2.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.14.w2.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.14.w2.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.15.v1.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.15.v1.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.15.v1.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.15.w1.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.15.w1.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.15.w1.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.15.w2.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.15.w2.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.15.w2.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.2.v1.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.2.v1.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.2.v1.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.2.w1.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.2.w1.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.2.w1.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.2.w2.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.2.w2.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.2.w2.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.3.v1.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.3.v1.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.3.v1.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.3.w1.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.3.w1.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.3.w1.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.3.w2.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.3.w2.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.3.w2.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.4.v1.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.4.v1.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.4.v1.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.4.w1.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.4.w1.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.4.w1.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.4.w2.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.4.w2.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.4.w2.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.5.v1.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.5.v1.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.5.v1.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.5.w1.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.5.w1.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.5.w1.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.5.w2.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.5.w2.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.5.w2.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.6.v1.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.6.v1.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.6.v1.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.6.w1.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.6.w1.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.6.w1.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.6.w2.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.6.w2.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.6.w2.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.7.v1.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.7.v1.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.7.v1.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.7.w1.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.7.w1.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.7.w1.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.7.w2.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.7.w2.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.7.w2.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.8.v1.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.8.v1.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.8.v1.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.8.w1.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.8.w1.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.8.w1.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.8.w2.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.8.w2.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.8.w2.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.9.v1.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.9.v1.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.9.v1.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.9.w1.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.9.w1.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.9.w1.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.9.w2.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.9.w2.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.experts.9.w2.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.router.layer.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.router.layer.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.ffn.router.layer.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.norm_attn_norm.attn.Wqkv.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.norm_attn_norm.attn.Wqkv.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.norm_attn_norm.attn.Wqkv.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.norm_attn_norm.attn.out_proj.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.norm_attn_norm.attn.out_proj.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.norm_attn_norm.attn.out_proj.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.norm_attn_norm.norm_1.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.17.norm_attn_norm.norm_2.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.0.v1.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.0.v1.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.0.v1.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.0.w1.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.0.w1.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.0.w1.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.0.w2.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.0.w2.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.0.w2.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.1.v1.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.1.v1.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.1.v1.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.1.w1.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.1.w1.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.1.w1.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.1.w2.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.1.w2.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.1.w2.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.10.v1.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.10.v1.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.10.v1.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.10.w1.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.10.w1.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.10.w1.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.10.w2.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.10.w2.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.10.w2.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.11.v1.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.11.v1.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.11.v1.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.11.w1.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.11.w1.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.11.w1.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.11.w2.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.11.w2.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.11.w2.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.12.v1.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.12.v1.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.12.v1.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.12.w1.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.12.w1.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.12.w1.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.12.w2.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.12.w2.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.12.w2.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.13.v1.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.13.v1.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.13.v1.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.13.w1.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.13.w1.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.13.w1.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.13.w2.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.13.w2.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.13.w2.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.14.v1.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.14.v1.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.14.v1.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.14.w1.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.14.w1.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.14.w1.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.14.w2.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.14.w2.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.14.w2.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.15.v1.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.15.v1.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.15.v1.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.15.w1.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.15.w1.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.15.w1.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.15.w2.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.15.w2.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.15.w2.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.2.v1.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.2.v1.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.2.v1.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.2.w1.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.2.w1.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.2.w1.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.2.w2.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.2.w2.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.2.w2.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.3.v1.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.3.v1.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.3.v1.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.3.w1.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.3.w1.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.3.w1.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.3.w2.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.3.w2.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.3.w2.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.4.v1.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.4.v1.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.4.v1.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.4.w1.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.4.w1.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.4.w1.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.4.w2.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.4.w2.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.4.w2.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.5.v1.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.5.v1.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.5.v1.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.5.w1.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.5.w1.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.5.w1.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.5.w2.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.5.w2.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.5.w2.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.6.v1.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.6.v1.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.6.v1.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.6.w1.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.6.w1.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.6.w1.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.6.w2.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.6.w2.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.6.w2.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.7.v1.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.7.v1.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.7.v1.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.7.w1.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.7.w1.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.7.w1.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.7.w2.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.7.w2.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.7.w2.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.8.v1.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.8.v1.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.8.v1.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.8.w1.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.8.w1.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.8.w1.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.8.w2.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.8.w2.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.8.w2.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.9.v1.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.9.v1.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.9.v1.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.9.w1.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.9.w1.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.9.w1.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.9.w2.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.9.w2.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.experts.9.w2.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.router.layer.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.router.layer.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.ffn.router.layer.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.norm_attn_norm.attn.Wqkv.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.norm_attn_norm.attn.Wqkv.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.norm_attn_norm.attn.Wqkv.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.norm_attn_norm.attn.out_proj.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.norm_attn_norm.attn.out_proj.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.norm_attn_norm.attn.out_proj.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.norm_attn_norm.norm_1.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.18.norm_attn_norm.norm_2.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.0.v1.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.0.v1.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.0.v1.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.0.w1.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.0.w1.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.0.w1.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.0.w2.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.0.w2.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.0.w2.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.1.v1.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.1.v1.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.1.v1.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.1.w1.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.1.w1.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.1.w1.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.1.w2.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.1.w2.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.1.w2.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.10.v1.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.10.v1.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.10.v1.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.10.w1.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.10.w1.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.10.w1.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.10.w2.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.10.w2.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.10.w2.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.11.v1.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.11.v1.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.11.v1.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.11.w1.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.11.w1.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.11.w1.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.11.w2.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.11.w2.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.11.w2.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.12.v1.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.12.v1.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.12.v1.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.12.w1.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.12.w1.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.12.w1.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.12.w2.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.12.w2.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.12.w2.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.13.v1.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.13.v1.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.13.v1.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.13.w1.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.13.w1.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.13.w1.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.13.w2.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.13.w2.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.13.w2.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.14.v1.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.14.v1.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.14.v1.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.14.w1.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.14.w1.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.14.w1.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.14.w2.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.14.w2.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.14.w2.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.15.v1.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.15.v1.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.15.v1.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.15.w1.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.15.w1.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.15.w1.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.15.w2.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.15.w2.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.15.w2.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.2.v1.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.2.v1.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.2.v1.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.2.w1.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.2.w1.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.2.w1.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.2.w2.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.2.w2.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.2.w2.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.3.v1.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.3.v1.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.3.v1.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.3.w1.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.3.w1.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.3.w1.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.3.w2.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.3.w2.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.3.w2.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.4.v1.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.4.v1.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.4.v1.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.4.w1.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.4.w1.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.4.w1.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.4.w2.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.4.w2.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.4.w2.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.5.v1.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.5.v1.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.5.v1.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.5.w1.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.5.w1.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.5.w1.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.5.w2.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.5.w2.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.5.w2.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.6.v1.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.6.v1.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.6.v1.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.6.w1.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.6.w1.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.6.w1.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.6.w2.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.6.w2.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.6.w2.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.7.v1.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.7.v1.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.7.v1.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.7.w1.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.7.w1.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.7.w1.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.7.w2.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.7.w2.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.7.w2.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.8.v1.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.8.v1.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.8.v1.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.8.w1.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.8.w1.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.8.w1.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.8.w2.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.8.w2.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.8.w2.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.9.v1.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.9.v1.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.9.v1.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.9.w1.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.9.w1.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.9.w1.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.9.w2.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.9.w2.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.experts.9.w2.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.router.layer.biases": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.router.layer.scales": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.ffn.router.layer.weight": "model-00007-of-00014.safetensors",
+        "transformer.blocks.19.norm_attn_norm.attn.Wqkv.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.19.norm_attn_norm.attn.Wqkv.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.19.norm_attn_norm.attn.Wqkv.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.19.norm_attn_norm.attn.out_proj.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.19.norm_attn_norm.attn.out_proj.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.19.norm_attn_norm.attn.out_proj.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.19.norm_attn_norm.norm_1.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.19.norm_attn_norm.norm_2.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.0.v1.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.0.v1.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.0.v1.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.0.w1.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.0.w1.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.0.w1.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.0.w2.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.0.w2.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.0.w2.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.1.v1.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.1.v1.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.1.v1.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.1.w1.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.1.w1.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.1.w1.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.1.w2.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.1.w2.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.1.w2.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.10.v1.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.10.v1.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.10.v1.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.10.w1.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.10.w1.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.10.w1.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.10.w2.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.10.w2.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.10.w2.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.11.v1.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.11.v1.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.11.v1.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.11.w1.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.11.w1.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.11.w1.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.11.w2.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.11.w2.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.11.w2.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.12.v1.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.12.v1.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.12.v1.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.12.w1.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.12.w1.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.12.w1.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.12.w2.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.12.w2.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.12.w2.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.13.v1.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.13.v1.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.13.v1.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.13.w1.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.13.w1.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.13.w1.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.13.w2.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.13.w2.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.13.w2.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.14.v1.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.14.v1.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.14.v1.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.14.w1.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.14.w1.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.14.w1.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.14.w2.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.14.w2.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.14.w2.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.15.v1.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.15.v1.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.15.v1.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.15.w1.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.15.w1.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.15.w1.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.15.w2.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.15.w2.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.15.w2.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.2.v1.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.2.v1.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.2.v1.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.2.w1.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.2.w1.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.2.w1.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.2.w2.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.2.w2.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.2.w2.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.3.v1.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.3.v1.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.3.v1.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.3.w1.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.3.w1.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.3.w1.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.3.w2.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.3.w2.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.3.w2.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.4.v1.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.4.v1.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.4.v1.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.4.w1.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.4.w1.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.4.w1.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.4.w2.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.4.w2.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.4.w2.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.5.v1.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.5.v1.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.5.v1.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.5.w1.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.5.w1.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.5.w1.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.5.w2.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.5.w2.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.5.w2.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.6.v1.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.6.v1.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.6.v1.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.6.w1.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.6.w1.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.6.w1.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.6.w2.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.6.w2.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.6.w2.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.7.v1.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.7.v1.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.7.v1.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.7.w1.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.7.w1.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.7.w1.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.7.w2.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.7.w2.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.7.w2.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.8.v1.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.8.v1.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.8.v1.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.8.w1.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.8.w1.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.8.w1.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.8.w2.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.8.w2.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.8.w2.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.9.v1.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.9.v1.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.9.v1.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.9.w1.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.9.w1.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.9.w1.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.9.w2.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.9.w2.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.experts.9.w2.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.ffn.router.layer.biases": "model-00001-of-00014.safetensors",
+        "transformer.blocks.2.ffn.router.layer.scales": "model-00001-of-00014.safetensors",
+        "transformer.blocks.2.ffn.router.layer.weight": "model-00001-of-00014.safetensors",
+        "transformer.blocks.2.norm_attn_norm.attn.Wqkv.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.norm_attn_norm.attn.Wqkv.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.norm_attn_norm.attn.Wqkv.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.norm_attn_norm.attn.out_proj.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.norm_attn_norm.attn.out_proj.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.norm_attn_norm.attn.out_proj.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.norm_attn_norm.norm_1.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.2.norm_attn_norm.norm_2.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.0.v1.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.0.v1.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.0.v1.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.0.w1.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.0.w1.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.0.w1.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.0.w2.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.0.w2.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.0.w2.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.1.v1.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.1.v1.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.1.v1.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.1.w1.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.1.w1.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.1.w1.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.1.w2.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.1.w2.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.1.w2.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.10.v1.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.10.v1.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.10.v1.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.10.w1.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.10.w1.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.10.w1.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.10.w2.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.10.w2.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.10.w2.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.11.v1.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.11.v1.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.11.v1.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.11.w1.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.11.w1.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.11.w1.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.11.w2.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.11.w2.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.11.w2.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.12.v1.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.12.v1.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.12.v1.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.12.w1.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.12.w1.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.12.w1.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.12.w2.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.12.w2.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.12.w2.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.13.v1.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.13.v1.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.13.v1.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.13.w1.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.13.w1.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.13.w1.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.13.w2.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.13.w2.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.13.w2.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.14.v1.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.14.v1.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.14.v1.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.14.w1.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.14.w1.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.14.w1.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.14.w2.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.14.w2.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.14.w2.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.15.v1.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.15.v1.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.15.v1.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.15.w1.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.15.w1.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.15.w1.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.15.w2.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.15.w2.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.15.w2.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.2.v1.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.2.v1.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.2.v1.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.2.w1.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.2.w1.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.2.w1.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.2.w2.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.2.w2.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.2.w2.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.3.v1.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.3.v1.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.3.v1.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.3.w1.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.3.w1.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.3.w1.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.3.w2.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.3.w2.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.3.w2.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.4.v1.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.4.v1.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.4.v1.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.4.w1.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.4.w1.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.4.w1.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.4.w2.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.4.w2.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.4.w2.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.5.v1.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.5.v1.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.5.v1.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.5.w1.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.5.w1.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.5.w1.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.5.w2.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.5.w2.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.5.w2.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.6.v1.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.6.v1.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.6.v1.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.6.w1.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.6.w1.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.6.w1.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.6.w2.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.6.w2.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.6.w2.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.7.v1.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.7.v1.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.7.v1.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.7.w1.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.7.w1.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.7.w1.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.7.w2.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.7.w2.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.7.w2.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.8.v1.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.8.v1.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.8.v1.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.8.w1.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.8.w1.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.8.w1.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.8.w2.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.8.w2.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.8.w2.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.9.v1.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.9.v1.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.9.v1.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.9.w1.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.9.w1.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.9.w1.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.9.w2.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.9.w2.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.experts.9.w2.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.router.layer.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.router.layer.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.ffn.router.layer.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.norm_attn_norm.attn.Wqkv.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.norm_attn_norm.attn.Wqkv.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.norm_attn_norm.attn.Wqkv.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.norm_attn_norm.attn.out_proj.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.norm_attn_norm.attn.out_proj.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.norm_attn_norm.attn.out_proj.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.norm_attn_norm.norm_1.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.20.norm_attn_norm.norm_2.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.0.v1.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.0.v1.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.0.v1.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.0.w1.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.0.w1.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.0.w1.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.0.w2.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.0.w2.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.0.w2.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.1.v1.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.1.v1.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.1.v1.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.1.w1.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.1.w1.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.1.w1.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.1.w2.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.1.w2.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.1.w2.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.10.v1.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.10.v1.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.10.v1.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.10.w1.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.10.w1.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.10.w1.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.10.w2.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.10.w2.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.10.w2.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.11.v1.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.11.v1.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.11.v1.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.11.w1.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.11.w1.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.11.w1.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.11.w2.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.11.w2.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.11.w2.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.12.v1.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.12.v1.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.12.v1.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.12.w1.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.12.w1.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.12.w1.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.12.w2.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.12.w2.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.12.w2.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.13.v1.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.13.v1.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.13.v1.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.13.w1.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.13.w1.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.13.w1.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.13.w2.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.13.w2.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.13.w2.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.14.v1.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.14.v1.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.14.v1.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.14.w1.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.14.w1.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.14.w1.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.14.w2.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.14.w2.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.14.w2.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.15.v1.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.15.v1.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.15.v1.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.15.w1.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.15.w1.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.15.w1.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.15.w2.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.15.w2.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.15.w2.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.2.v1.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.2.v1.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.2.v1.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.2.w1.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.2.w1.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.2.w1.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.2.w2.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.2.w2.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.2.w2.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.3.v1.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.3.v1.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.3.v1.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.3.w1.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.3.w1.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.3.w1.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.3.w2.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.3.w2.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.3.w2.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.4.v1.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.4.v1.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.4.v1.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.4.w1.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.4.w1.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.4.w1.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.4.w2.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.4.w2.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.4.w2.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.5.v1.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.5.v1.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.5.v1.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.5.w1.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.5.w1.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.5.w1.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.5.w2.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.5.w2.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.5.w2.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.6.v1.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.6.v1.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.6.v1.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.6.w1.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.6.w1.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.6.w1.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.6.w2.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.6.w2.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.6.w2.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.7.v1.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.7.v1.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.7.v1.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.7.w1.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.7.w1.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.7.w1.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.7.w2.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.7.w2.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.7.w2.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.8.v1.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.8.v1.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.8.v1.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.8.w1.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.8.w1.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.8.w1.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.8.w2.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.8.w2.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.8.w2.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.9.v1.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.9.v1.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.9.v1.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.9.w1.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.9.w1.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.9.w1.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.9.w2.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.9.w2.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.experts.9.w2.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.router.layer.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.router.layer.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.ffn.router.layer.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.norm_attn_norm.attn.Wqkv.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.norm_attn_norm.attn.Wqkv.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.norm_attn_norm.attn.Wqkv.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.norm_attn_norm.attn.out_proj.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.norm_attn_norm.attn.out_proj.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.norm_attn_norm.attn.out_proj.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.norm_attn_norm.norm_1.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.21.norm_attn_norm.norm_2.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.0.v1.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.0.v1.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.0.v1.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.0.w1.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.0.w1.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.0.w1.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.0.w2.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.0.w2.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.0.w2.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.1.v1.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.1.v1.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.1.v1.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.1.w1.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.1.w1.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.1.w1.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.1.w2.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.1.w2.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.1.w2.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.10.v1.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.10.v1.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.10.v1.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.10.w1.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.10.w1.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.10.w1.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.10.w2.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.10.w2.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.10.w2.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.11.v1.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.11.v1.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.11.v1.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.11.w1.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.11.w1.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.11.w1.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.11.w2.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.11.w2.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.11.w2.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.12.v1.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.12.v1.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.12.v1.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.12.w1.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.12.w1.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.12.w1.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.12.w2.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.12.w2.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.12.w2.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.13.v1.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.13.v1.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.13.v1.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.13.w1.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.13.w1.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.13.w1.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.13.w2.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.13.w2.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.13.w2.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.14.v1.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.14.v1.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.14.v1.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.14.w1.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.14.w1.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.14.w1.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.14.w2.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.14.w2.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.14.w2.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.15.v1.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.15.v1.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.15.v1.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.15.w1.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.15.w1.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.15.w1.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.15.w2.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.15.w2.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.15.w2.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.2.v1.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.2.v1.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.2.v1.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.2.w1.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.2.w1.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.2.w1.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.2.w2.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.2.w2.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.2.w2.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.3.v1.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.3.v1.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.3.v1.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.3.w1.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.3.w1.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.3.w1.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.3.w2.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.3.w2.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.3.w2.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.4.v1.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.4.v1.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.4.v1.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.4.w1.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.4.w1.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.4.w1.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.4.w2.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.4.w2.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.4.w2.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.5.v1.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.5.v1.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.5.v1.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.5.w1.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.5.w1.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.5.w1.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.5.w2.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.5.w2.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.5.w2.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.6.v1.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.6.v1.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.6.v1.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.6.w1.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.6.w1.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.6.w1.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.6.w2.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.6.w2.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.6.w2.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.7.v1.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.7.v1.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.7.v1.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.7.w1.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.7.w1.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.7.w1.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.7.w2.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.7.w2.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.7.w2.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.8.v1.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.8.v1.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.8.v1.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.8.w1.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.8.w1.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.8.w1.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.8.w2.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.8.w2.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.8.w2.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.9.v1.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.9.v1.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.9.v1.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.9.w1.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.9.w1.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.9.w1.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.9.w2.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.9.w2.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.experts.9.w2.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.router.layer.biases": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.router.layer.scales": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.ffn.router.layer.weight": "model-00008-of-00014.safetensors",
+        "transformer.blocks.22.norm_attn_norm.attn.Wqkv.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.22.norm_attn_norm.attn.Wqkv.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.22.norm_attn_norm.attn.Wqkv.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.22.norm_attn_norm.attn.out_proj.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.22.norm_attn_norm.attn.out_proj.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.22.norm_attn_norm.attn.out_proj.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.22.norm_attn_norm.norm_1.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.22.norm_attn_norm.norm_2.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.0.v1.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.0.v1.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.0.v1.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.0.w1.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.0.w1.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.0.w1.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.0.w2.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.0.w2.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.0.w2.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.1.v1.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.1.v1.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.1.v1.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.1.w1.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.1.w1.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.1.w1.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.1.w2.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.1.w2.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.1.w2.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.10.v1.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.10.v1.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.10.v1.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.10.w1.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.10.w1.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.10.w1.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.10.w2.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.10.w2.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.10.w2.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.11.v1.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.11.v1.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.11.v1.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.11.w1.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.11.w1.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.11.w1.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.11.w2.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.11.w2.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.11.w2.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.12.v1.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.12.v1.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.12.v1.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.12.w1.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.12.w1.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.12.w1.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.12.w2.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.12.w2.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.12.w2.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.13.v1.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.13.v1.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.13.v1.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.13.w1.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.13.w1.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.13.w1.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.13.w2.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.13.w2.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.13.w2.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.14.v1.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.14.v1.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.14.v1.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.14.w1.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.14.w1.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.14.w1.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.14.w2.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.14.w2.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.14.w2.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.15.v1.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.15.v1.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.15.v1.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.15.w1.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.15.w1.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.15.w1.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.15.w2.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.15.w2.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.15.w2.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.2.v1.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.2.v1.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.2.v1.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.2.w1.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.2.w1.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.2.w1.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.2.w2.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.2.w2.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.2.w2.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.3.v1.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.3.v1.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.3.v1.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.3.w1.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.3.w1.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.3.w1.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.3.w2.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.3.w2.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.3.w2.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.4.v1.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.4.v1.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.4.v1.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.4.w1.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.4.w1.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.4.w1.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.4.w2.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.4.w2.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.4.w2.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.5.v1.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.5.v1.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.5.v1.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.5.w1.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.5.w1.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.5.w1.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.5.w2.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.5.w2.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.5.w2.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.6.v1.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.6.v1.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.6.v1.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.6.w1.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.6.w1.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.6.w1.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.6.w2.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.6.w2.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.6.w2.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.7.v1.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.7.v1.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.7.v1.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.7.w1.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.7.w1.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.7.w1.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.7.w2.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.7.w2.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.7.w2.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.8.v1.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.8.v1.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.8.v1.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.8.w1.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.8.w1.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.8.w1.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.8.w2.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.8.w2.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.8.w2.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.9.v1.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.9.v1.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.9.v1.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.9.w1.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.9.w1.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.9.w1.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.9.w2.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.9.w2.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.experts.9.w2.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.router.layer.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.router.layer.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.ffn.router.layer.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.norm_attn_norm.attn.Wqkv.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.norm_attn_norm.attn.Wqkv.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.norm_attn_norm.attn.Wqkv.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.norm_attn_norm.attn.out_proj.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.norm_attn_norm.attn.out_proj.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.norm_attn_norm.attn.out_proj.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.norm_attn_norm.norm_1.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.23.norm_attn_norm.norm_2.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.0.v1.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.0.v1.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.0.v1.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.0.w1.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.0.w1.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.0.w1.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.0.w2.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.0.w2.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.0.w2.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.1.v1.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.1.v1.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.1.v1.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.1.w1.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.1.w1.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.1.w1.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.1.w2.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.1.w2.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.1.w2.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.10.v1.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.10.v1.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.10.v1.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.10.w1.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.10.w1.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.10.w1.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.10.w2.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.10.w2.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.10.w2.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.11.v1.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.11.v1.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.11.v1.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.11.w1.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.11.w1.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.11.w1.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.11.w2.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.11.w2.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.11.w2.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.12.v1.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.12.v1.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.12.v1.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.12.w1.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.12.w1.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.12.w1.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.12.w2.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.12.w2.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.12.w2.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.13.v1.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.13.v1.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.13.v1.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.13.w1.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.13.w1.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.13.w1.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.13.w2.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.13.w2.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.13.w2.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.14.v1.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.14.v1.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.14.v1.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.14.w1.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.14.w1.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.14.w1.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.14.w2.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.14.w2.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.14.w2.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.15.v1.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.15.v1.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.15.v1.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.15.w1.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.15.w1.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.15.w1.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.15.w2.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.15.w2.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.15.w2.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.2.v1.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.2.v1.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.2.v1.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.2.w1.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.2.w1.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.2.w1.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.2.w2.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.2.w2.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.2.w2.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.3.v1.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.3.v1.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.3.v1.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.3.w1.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.3.w1.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.3.w1.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.3.w2.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.3.w2.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.3.w2.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.4.v1.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.4.v1.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.4.v1.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.4.w1.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.4.w1.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.4.w1.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.4.w2.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.4.w2.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.4.w2.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.5.v1.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.5.v1.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.5.v1.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.5.w1.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.5.w1.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.5.w1.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.5.w2.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.5.w2.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.5.w2.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.6.v1.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.6.v1.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.6.v1.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.6.w1.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.6.w1.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.6.w1.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.6.w2.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.6.w2.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.6.w2.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.7.v1.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.7.v1.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.7.v1.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.7.w1.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.7.w1.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.7.w1.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.7.w2.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.7.w2.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.7.w2.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.8.v1.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.8.v1.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.8.v1.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.8.w1.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.8.w1.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.8.w1.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.8.w2.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.8.w2.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.8.w2.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.9.v1.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.9.v1.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.9.v1.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.9.w1.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.9.w1.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.9.w1.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.9.w2.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.9.w2.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.experts.9.w2.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.router.layer.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.router.layer.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.ffn.router.layer.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.norm_attn_norm.attn.Wqkv.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.norm_attn_norm.attn.Wqkv.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.norm_attn_norm.attn.Wqkv.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.norm_attn_norm.attn.out_proj.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.norm_attn_norm.attn.out_proj.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.norm_attn_norm.attn.out_proj.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.norm_attn_norm.norm_1.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.24.norm_attn_norm.norm_2.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.0.v1.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.0.v1.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.0.v1.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.0.w1.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.0.w1.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.0.w1.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.0.w2.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.0.w2.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.0.w2.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.1.v1.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.1.v1.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.1.v1.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.1.w1.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.1.w1.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.1.w1.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.1.w2.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.1.w2.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.1.w2.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.10.v1.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.10.v1.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.10.v1.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.10.w1.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.10.w1.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.10.w1.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.10.w2.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.10.w2.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.10.w2.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.11.v1.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.11.v1.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.11.v1.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.11.w1.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.11.w1.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.11.w1.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.11.w2.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.11.w2.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.11.w2.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.12.v1.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.12.v1.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.12.v1.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.12.w1.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.12.w1.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.12.w1.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.12.w2.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.12.w2.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.12.w2.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.13.v1.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.13.v1.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.13.v1.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.13.w1.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.13.w1.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.13.w1.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.13.w2.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.13.w2.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.13.w2.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.14.v1.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.14.v1.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.14.v1.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.14.w1.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.14.w1.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.14.w1.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.14.w2.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.14.w2.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.14.w2.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.15.v1.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.15.v1.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.15.v1.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.15.w1.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.15.w1.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.15.w1.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.15.w2.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.15.w2.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.15.w2.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.2.v1.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.2.v1.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.2.v1.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.2.w1.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.2.w1.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.2.w1.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.2.w2.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.2.w2.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.2.w2.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.3.v1.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.3.v1.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.3.v1.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.3.w1.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.3.w1.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.3.w1.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.3.w2.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.3.w2.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.3.w2.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.4.v1.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.4.v1.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.4.v1.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.4.w1.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.4.w1.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.4.w1.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.4.w2.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.4.w2.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.4.w2.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.5.v1.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.5.v1.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.5.v1.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.5.w1.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.5.w1.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.5.w1.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.5.w2.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.5.w2.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.5.w2.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.6.v1.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.6.v1.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.6.v1.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.6.w1.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.6.w1.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.6.w1.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.6.w2.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.6.w2.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.6.w2.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.7.v1.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.7.v1.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.7.v1.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.7.w1.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.7.w1.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.7.w1.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.7.w2.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.7.w2.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.7.w2.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.8.v1.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.8.v1.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.8.v1.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.8.w1.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.8.w1.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.8.w1.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.8.w2.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.8.w2.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.8.w2.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.9.v1.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.9.v1.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.9.v1.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.9.w1.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.9.w1.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.9.w1.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.9.w2.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.9.w2.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.25.ffn.experts.9.w2.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.25.ffn.router.layer.biases": "model-00009-of-00014.safetensors",
+        "transformer.blocks.25.ffn.router.layer.scales": "model-00009-of-00014.safetensors",
+        "transformer.blocks.25.ffn.router.layer.weight": "model-00009-of-00014.safetensors",
+        "transformer.blocks.25.norm_attn_norm.attn.Wqkv.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.25.norm_attn_norm.attn.Wqkv.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.25.norm_attn_norm.attn.Wqkv.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.25.norm_attn_norm.attn.out_proj.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.25.norm_attn_norm.attn.out_proj.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.25.norm_attn_norm.attn.out_proj.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.25.norm_attn_norm.norm_1.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.25.norm_attn_norm.norm_2.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.0.v1.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.0.v1.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.0.v1.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.0.w1.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.0.w1.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.0.w1.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.0.w2.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.0.w2.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.0.w2.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.1.v1.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.1.v1.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.1.v1.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.1.w1.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.1.w1.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.1.w1.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.1.w2.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.1.w2.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.1.w2.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.10.v1.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.10.v1.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.10.v1.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.10.w1.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.10.w1.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.10.w1.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.10.w2.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.10.w2.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.10.w2.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.11.v1.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.11.v1.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.11.v1.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.11.w1.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.11.w1.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.11.w1.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.11.w2.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.11.w2.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.11.w2.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.12.v1.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.12.v1.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.12.v1.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.12.w1.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.12.w1.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.12.w1.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.12.w2.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.12.w2.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.12.w2.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.13.v1.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.13.v1.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.13.v1.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.13.w1.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.13.w1.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.13.w1.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.13.w2.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.13.w2.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.13.w2.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.14.v1.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.14.v1.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.14.v1.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.14.w1.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.14.w1.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.14.w1.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.14.w2.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.14.w2.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.14.w2.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.15.v1.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.15.v1.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.15.v1.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.15.w1.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.15.w1.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.15.w1.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.15.w2.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.15.w2.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.15.w2.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.2.v1.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.2.v1.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.2.v1.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.2.w1.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.2.w1.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.2.w1.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.2.w2.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.2.w2.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.2.w2.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.3.v1.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.3.v1.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.3.v1.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.3.w1.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.3.w1.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.3.w1.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.3.w2.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.3.w2.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.3.w2.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.4.v1.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.4.v1.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.4.v1.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.4.w1.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.4.w1.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.4.w1.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.4.w2.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.4.w2.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.4.w2.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.5.v1.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.5.v1.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.5.v1.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.5.w1.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.5.w1.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.5.w1.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.5.w2.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.5.w2.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.5.w2.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.6.v1.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.6.v1.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.6.v1.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.6.w1.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.6.w1.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.6.w1.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.6.w2.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.6.w2.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.6.w2.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.7.v1.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.7.v1.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.7.v1.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.7.w1.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.7.w1.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.7.w1.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.7.w2.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.7.w2.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.7.w2.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.8.v1.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.8.v1.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.8.v1.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.8.w1.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.8.w1.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.8.w1.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.8.w2.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.8.w2.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.8.w2.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.9.v1.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.9.v1.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.9.v1.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.9.w1.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.9.w1.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.9.w1.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.9.w2.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.9.w2.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.experts.9.w2.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.router.layer.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.router.layer.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.ffn.router.layer.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.norm_attn_norm.attn.Wqkv.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.norm_attn_norm.attn.Wqkv.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.norm_attn_norm.attn.Wqkv.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.norm_attn_norm.attn.out_proj.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.norm_attn_norm.attn.out_proj.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.norm_attn_norm.attn.out_proj.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.norm_attn_norm.norm_1.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.26.norm_attn_norm.norm_2.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.0.v1.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.0.v1.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.0.v1.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.0.w1.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.0.w1.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.0.w1.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.0.w2.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.0.w2.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.0.w2.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.1.v1.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.1.v1.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.1.v1.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.1.w1.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.1.w1.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.1.w1.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.1.w2.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.1.w2.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.1.w2.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.10.v1.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.10.v1.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.10.v1.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.10.w1.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.10.w1.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.10.w1.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.10.w2.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.10.w2.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.10.w2.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.11.v1.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.11.v1.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.11.v1.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.11.w1.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.11.w1.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.11.w1.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.11.w2.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.11.w2.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.11.w2.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.12.v1.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.12.v1.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.12.v1.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.12.w1.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.12.w1.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.12.w1.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.12.w2.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.12.w2.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.12.w2.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.13.v1.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.13.v1.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.13.v1.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.13.w1.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.13.w1.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.13.w1.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.13.w2.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.13.w2.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.13.w2.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.14.v1.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.14.v1.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.14.v1.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.14.w1.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.14.w1.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.14.w1.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.14.w2.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.14.w2.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.14.w2.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.15.v1.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.15.v1.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.15.v1.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.15.w1.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.15.w1.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.15.w1.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.15.w2.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.15.w2.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.15.w2.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.2.v1.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.2.v1.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.2.v1.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.2.w1.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.2.w1.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.2.w1.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.2.w2.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.2.w2.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.2.w2.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.3.v1.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.3.v1.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.3.v1.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.3.w1.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.3.w1.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.3.w1.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.3.w2.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.3.w2.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.3.w2.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.4.v1.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.4.v1.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.4.v1.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.4.w1.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.4.w1.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.4.w1.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.4.w2.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.4.w2.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.4.w2.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.5.v1.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.5.v1.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.5.v1.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.5.w1.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.5.w1.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.5.w1.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.5.w2.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.5.w2.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.5.w2.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.6.v1.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.6.v1.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.6.v1.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.6.w1.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.6.w1.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.6.w1.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.6.w2.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.6.w2.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.6.w2.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.7.v1.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.7.v1.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.7.v1.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.7.w1.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.7.w1.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.7.w1.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.7.w2.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.7.w2.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.7.w2.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.8.v1.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.8.v1.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.8.v1.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.8.w1.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.8.w1.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.8.w1.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.8.w2.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.8.w2.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.8.w2.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.9.v1.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.9.v1.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.9.v1.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.9.w1.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.9.w1.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.9.w1.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.9.w2.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.9.w2.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.experts.9.w2.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.router.layer.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.router.layer.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.ffn.router.layer.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.norm_attn_norm.attn.Wqkv.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.norm_attn_norm.attn.Wqkv.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.norm_attn_norm.attn.Wqkv.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.norm_attn_norm.attn.out_proj.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.norm_attn_norm.attn.out_proj.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.norm_attn_norm.attn.out_proj.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.norm_attn_norm.norm_1.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.27.norm_attn_norm.norm_2.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.0.v1.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.0.v1.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.0.v1.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.0.w1.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.0.w1.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.0.w1.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.0.w2.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.0.w2.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.0.w2.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.1.v1.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.1.v1.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.1.v1.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.1.w1.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.1.w1.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.1.w1.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.1.w2.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.1.w2.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.1.w2.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.10.v1.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.10.v1.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.10.v1.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.10.w1.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.10.w1.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.10.w1.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.10.w2.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.10.w2.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.10.w2.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.11.v1.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.11.v1.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.11.v1.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.11.w1.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.11.w1.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.11.w1.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.11.w2.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.11.w2.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.11.w2.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.12.v1.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.12.v1.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.12.v1.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.12.w1.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.12.w1.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.12.w1.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.12.w2.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.12.w2.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.12.w2.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.13.v1.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.13.v1.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.13.v1.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.13.w1.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.13.w1.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.13.w1.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.13.w2.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.13.w2.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.13.w2.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.14.v1.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.14.v1.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.14.v1.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.14.w1.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.14.w1.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.14.w1.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.14.w2.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.14.w2.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.14.w2.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.15.v1.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.15.v1.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.15.v1.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.15.w1.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.15.w1.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.15.w1.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.15.w2.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.15.w2.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.15.w2.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.2.v1.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.2.v1.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.2.v1.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.2.w1.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.2.w1.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.2.w1.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.2.w2.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.2.w2.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.2.w2.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.3.v1.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.3.v1.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.3.v1.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.3.w1.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.3.w1.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.3.w1.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.3.w2.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.3.w2.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.3.w2.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.4.v1.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.4.v1.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.4.v1.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.4.w1.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.4.w1.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.4.w1.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.4.w2.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.4.w2.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.4.w2.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.5.v1.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.5.v1.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.5.v1.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.5.w1.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.5.w1.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.5.w1.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.5.w2.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.5.w2.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.5.w2.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.6.v1.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.6.v1.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.6.v1.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.6.w1.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.6.w1.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.6.w1.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.6.w2.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.6.w2.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.6.w2.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.7.v1.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.7.v1.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.7.v1.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.7.w1.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.7.w1.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.7.w1.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.7.w2.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.7.w2.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.7.w2.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.8.v1.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.8.v1.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.8.v1.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.8.w1.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.8.w1.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.8.w1.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.8.w2.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.8.w2.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.8.w2.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.9.v1.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.9.v1.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.9.v1.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.9.w1.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.9.w1.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.9.w1.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.9.w2.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.9.w2.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.28.ffn.experts.9.w2.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.28.ffn.router.layer.biases": "model-00010-of-00014.safetensors",
+        "transformer.blocks.28.ffn.router.layer.scales": "model-00010-of-00014.safetensors",
+        "transformer.blocks.28.ffn.router.layer.weight": "model-00010-of-00014.safetensors",
+        "transformer.blocks.28.norm_attn_norm.attn.Wqkv.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.28.norm_attn_norm.attn.Wqkv.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.28.norm_attn_norm.attn.Wqkv.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.28.norm_attn_norm.attn.out_proj.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.28.norm_attn_norm.attn.out_proj.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.28.norm_attn_norm.attn.out_proj.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.28.norm_attn_norm.norm_1.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.28.norm_attn_norm.norm_2.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.0.v1.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.0.v1.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.0.v1.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.0.w1.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.0.w1.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.0.w1.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.0.w2.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.0.w2.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.0.w2.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.1.v1.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.1.v1.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.1.v1.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.1.w1.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.1.w1.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.1.w1.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.1.w2.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.1.w2.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.1.w2.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.10.v1.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.10.v1.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.10.v1.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.10.w1.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.10.w1.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.10.w1.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.10.w2.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.10.w2.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.10.w2.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.11.v1.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.11.v1.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.11.v1.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.11.w1.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.11.w1.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.11.w1.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.11.w2.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.11.w2.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.11.w2.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.12.v1.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.12.v1.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.12.v1.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.12.w1.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.12.w1.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.12.w1.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.12.w2.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.12.w2.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.12.w2.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.13.v1.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.13.v1.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.13.v1.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.13.w1.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.13.w1.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.13.w1.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.13.w2.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.13.w2.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.13.w2.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.14.v1.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.14.v1.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.14.v1.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.14.w1.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.14.w1.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.14.w1.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.14.w2.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.14.w2.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.14.w2.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.15.v1.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.15.v1.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.15.v1.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.15.w1.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.15.w1.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.15.w1.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.15.w2.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.15.w2.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.15.w2.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.2.v1.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.2.v1.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.2.v1.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.2.w1.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.2.w1.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.2.w1.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.2.w2.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.2.w2.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.2.w2.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.3.v1.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.3.v1.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.3.v1.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.3.w1.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.3.w1.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.3.w1.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.3.w2.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.3.w2.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.3.w2.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.4.v1.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.4.v1.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.4.v1.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.4.w1.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.4.w1.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.4.w1.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.4.w2.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.4.w2.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.4.w2.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.5.v1.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.5.v1.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.5.v1.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.5.w1.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.5.w1.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.5.w1.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.5.w2.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.5.w2.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.5.w2.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.6.v1.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.6.v1.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.6.v1.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.6.w1.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.6.w1.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.6.w1.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.6.w2.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.6.w2.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.6.w2.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.7.v1.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.7.v1.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.7.v1.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.7.w1.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.7.w1.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.7.w1.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.7.w2.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.7.w2.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.7.w2.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.8.v1.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.8.v1.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.8.v1.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.8.w1.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.8.w1.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.8.w1.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.8.w2.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.8.w2.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.8.w2.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.9.v1.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.9.v1.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.9.v1.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.9.w1.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.9.w1.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.9.w1.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.9.w2.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.9.w2.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.experts.9.w2.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.router.layer.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.router.layer.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.ffn.router.layer.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.norm_attn_norm.attn.Wqkv.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.norm_attn_norm.attn.Wqkv.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.norm_attn_norm.attn.Wqkv.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.norm_attn_norm.attn.out_proj.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.norm_attn_norm.attn.out_proj.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.norm_attn_norm.attn.out_proj.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.norm_attn_norm.norm_1.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.29.norm_attn_norm.norm_2.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.0.v1.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.0.v1.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.0.v1.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.0.w1.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.0.w1.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.0.w1.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.0.w2.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.0.w2.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.0.w2.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.1.v1.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.1.v1.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.1.v1.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.1.w1.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.1.w1.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.1.w1.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.1.w2.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.1.w2.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.1.w2.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.10.v1.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.10.v1.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.10.v1.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.10.w1.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.10.w1.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.10.w1.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.10.w2.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.10.w2.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.10.w2.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.11.v1.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.11.v1.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.11.v1.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.11.w1.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.11.w1.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.11.w1.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.11.w2.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.11.w2.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.11.w2.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.12.v1.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.12.v1.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.12.v1.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.12.w1.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.12.w1.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.12.w1.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.12.w2.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.12.w2.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.12.w2.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.13.v1.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.13.v1.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.13.v1.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.13.w1.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.13.w1.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.13.w1.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.13.w2.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.13.w2.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.13.w2.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.14.v1.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.14.v1.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.14.v1.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.14.w1.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.14.w1.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.14.w1.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.14.w2.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.14.w2.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.14.w2.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.15.v1.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.15.v1.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.15.v1.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.15.w1.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.15.w1.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.15.w1.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.15.w2.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.15.w2.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.15.w2.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.2.v1.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.2.v1.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.2.v1.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.2.w1.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.2.w1.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.2.w1.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.2.w2.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.2.w2.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.2.w2.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.3.v1.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.3.v1.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.3.v1.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.3.w1.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.3.w1.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.3.w1.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.3.w2.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.3.w2.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.3.w2.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.4.v1.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.4.v1.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.4.v1.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.4.w1.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.4.w1.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.4.w1.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.4.w2.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.4.w2.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.4.w2.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.5.v1.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.5.v1.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.5.v1.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.5.w1.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.5.w1.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.5.w1.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.5.w2.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.5.w2.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.5.w2.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.6.v1.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.6.v1.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.6.v1.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.6.w1.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.6.w1.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.6.w1.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.6.w2.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.6.w2.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.6.w2.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.7.v1.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.7.v1.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.7.v1.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.7.w1.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.7.w1.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.7.w1.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.7.w2.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.7.w2.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.7.w2.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.8.v1.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.8.v1.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.8.v1.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.8.w1.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.8.w1.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.8.w1.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.8.w2.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.8.w2.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.8.w2.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.9.v1.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.9.v1.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.9.v1.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.9.w1.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.9.w1.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.9.w1.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.9.w2.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.9.w2.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.experts.9.w2.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.router.layer.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.router.layer.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.ffn.router.layer.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.norm_attn_norm.attn.Wqkv.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.norm_attn_norm.attn.Wqkv.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.norm_attn_norm.attn.Wqkv.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.norm_attn_norm.attn.out_proj.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.norm_attn_norm.attn.out_proj.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.norm_attn_norm.attn.out_proj.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.norm_attn_norm.norm_1.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.3.norm_attn_norm.norm_2.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.0.v1.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.0.v1.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.0.v1.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.0.w1.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.0.w1.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.0.w1.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.0.w2.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.0.w2.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.0.w2.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.1.v1.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.1.v1.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.1.v1.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.1.w1.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.1.w1.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.1.w1.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.1.w2.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.1.w2.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.1.w2.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.10.v1.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.10.v1.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.10.v1.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.10.w1.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.10.w1.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.10.w1.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.10.w2.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.10.w2.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.10.w2.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.11.v1.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.11.v1.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.11.v1.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.11.w1.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.11.w1.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.11.w1.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.11.w2.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.11.w2.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.11.w2.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.12.v1.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.12.v1.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.12.v1.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.12.w1.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.12.w1.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.12.w1.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.12.w2.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.12.w2.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.12.w2.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.13.v1.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.13.v1.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.13.v1.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.13.w1.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.13.w1.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.13.w1.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.13.w2.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.13.w2.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.13.w2.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.14.v1.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.14.v1.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.14.v1.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.14.w1.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.14.w1.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.14.w1.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.14.w2.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.14.w2.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.14.w2.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.15.v1.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.15.v1.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.15.v1.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.15.w1.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.15.w1.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.15.w1.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.15.w2.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.15.w2.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.15.w2.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.2.v1.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.2.v1.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.2.v1.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.2.w1.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.2.w1.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.2.w1.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.2.w2.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.2.w2.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.2.w2.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.3.v1.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.3.v1.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.3.v1.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.3.w1.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.3.w1.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.3.w1.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.3.w2.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.3.w2.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.3.w2.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.4.v1.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.4.v1.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.4.v1.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.4.w1.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.4.w1.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.4.w1.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.4.w2.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.4.w2.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.4.w2.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.5.v1.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.5.v1.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.5.v1.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.5.w1.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.5.w1.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.5.w1.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.5.w2.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.5.w2.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.5.w2.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.6.v1.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.6.v1.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.6.v1.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.6.w1.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.6.w1.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.6.w1.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.6.w2.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.6.w2.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.6.w2.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.7.v1.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.7.v1.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.7.v1.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.7.w1.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.7.w1.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.7.w1.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.7.w2.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.7.w2.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.7.w2.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.8.v1.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.8.v1.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.8.v1.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.8.w1.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.8.w1.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.8.w1.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.8.w2.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.8.w2.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.8.w2.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.9.v1.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.9.v1.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.9.v1.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.9.w1.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.9.w1.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.9.w1.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.9.w2.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.9.w2.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.experts.9.w2.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.router.layer.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.router.layer.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.ffn.router.layer.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.norm_attn_norm.attn.Wqkv.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.norm_attn_norm.attn.Wqkv.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.norm_attn_norm.attn.Wqkv.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.norm_attn_norm.attn.out_proj.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.norm_attn_norm.attn.out_proj.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.norm_attn_norm.attn.out_proj.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.norm_attn_norm.norm_1.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.30.norm_attn_norm.norm_2.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.0.v1.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.0.v1.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.0.v1.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.0.w1.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.0.w1.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.0.w1.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.0.w2.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.0.w2.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.0.w2.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.1.v1.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.1.v1.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.1.v1.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.1.w1.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.1.w1.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.1.w1.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.1.w2.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.1.w2.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.1.w2.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.10.v1.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.10.v1.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.10.v1.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.10.w1.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.10.w1.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.10.w1.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.10.w2.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.10.w2.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.10.w2.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.11.v1.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.11.v1.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.11.v1.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.11.w1.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.11.w1.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.11.w1.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.11.w2.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.11.w2.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.11.w2.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.12.v1.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.12.v1.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.12.v1.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.12.w1.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.12.w1.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.12.w1.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.12.w2.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.12.w2.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.12.w2.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.13.v1.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.13.v1.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.13.v1.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.13.w1.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.13.w1.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.13.w1.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.13.w2.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.13.w2.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.13.w2.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.14.v1.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.14.v1.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.14.v1.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.14.w1.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.14.w1.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.14.w1.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.14.w2.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.14.w2.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.14.w2.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.15.v1.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.15.v1.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.15.v1.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.15.w1.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.15.w1.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.15.w1.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.15.w2.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.15.w2.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.15.w2.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.2.v1.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.2.v1.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.2.v1.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.2.w1.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.2.w1.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.2.w1.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.2.w2.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.2.w2.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.2.w2.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.3.v1.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.3.v1.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.3.v1.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.3.w1.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.3.w1.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.3.w1.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.3.w2.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.3.w2.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.3.w2.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.4.v1.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.4.v1.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.4.v1.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.4.w1.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.4.w1.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.4.w1.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.4.w2.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.4.w2.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.4.w2.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.5.v1.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.5.v1.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.5.v1.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.5.w1.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.5.w1.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.5.w1.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.5.w2.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.5.w2.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.5.w2.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.6.v1.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.6.v1.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.6.v1.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.6.w1.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.6.w1.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.6.w1.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.6.w2.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.6.w2.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.6.w2.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.7.v1.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.7.v1.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.7.v1.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.7.w1.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.7.w1.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.7.w1.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.7.w2.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.7.w2.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.7.w2.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.8.v1.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.8.v1.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.8.v1.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.8.w1.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.8.w1.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.8.w1.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.8.w2.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.8.w2.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.8.w2.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.9.v1.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.9.v1.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.9.v1.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.9.w1.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.9.w1.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.9.w1.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.9.w2.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.9.w2.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.31.ffn.experts.9.w2.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.31.ffn.router.layer.biases": "model-00011-of-00014.safetensors",
+        "transformer.blocks.31.ffn.router.layer.scales": "model-00011-of-00014.safetensors",
+        "transformer.blocks.31.ffn.router.layer.weight": "model-00011-of-00014.safetensors",
+        "transformer.blocks.31.norm_attn_norm.attn.Wqkv.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.31.norm_attn_norm.attn.Wqkv.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.31.norm_attn_norm.attn.Wqkv.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.31.norm_attn_norm.attn.out_proj.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.31.norm_attn_norm.attn.out_proj.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.31.norm_attn_norm.attn.out_proj.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.31.norm_attn_norm.norm_1.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.31.norm_attn_norm.norm_2.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.0.v1.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.0.v1.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.0.v1.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.0.w1.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.0.w1.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.0.w1.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.0.w2.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.0.w2.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.0.w2.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.1.v1.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.1.v1.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.1.v1.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.1.w1.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.1.w1.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.1.w1.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.1.w2.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.1.w2.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.1.w2.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.10.v1.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.10.v1.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.10.v1.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.10.w1.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.10.w1.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.10.w1.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.10.w2.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.10.w2.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.10.w2.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.11.v1.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.11.v1.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.11.v1.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.11.w1.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.11.w1.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.11.w1.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.11.w2.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.11.w2.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.11.w2.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.12.v1.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.12.v1.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.12.v1.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.12.w1.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.12.w1.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.12.w1.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.12.w2.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.12.w2.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.12.w2.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.13.v1.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.13.v1.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.13.v1.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.13.w1.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.13.w1.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.13.w1.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.13.w2.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.13.w2.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.13.w2.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.14.v1.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.14.v1.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.14.v1.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.14.w1.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.14.w1.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.14.w1.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.14.w2.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.14.w2.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.14.w2.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.15.v1.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.15.v1.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.15.v1.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.15.w1.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.15.w1.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.15.w1.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.15.w2.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.15.w2.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.15.w2.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.2.v1.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.2.v1.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.2.v1.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.2.w1.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.2.w1.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.2.w1.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.2.w2.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.2.w2.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.2.w2.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.3.v1.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.3.v1.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.3.v1.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.3.w1.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.3.w1.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.3.w1.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.3.w2.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.3.w2.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.3.w2.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.4.v1.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.4.v1.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.4.v1.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.4.w1.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.4.w1.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.4.w1.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.4.w2.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.4.w2.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.4.w2.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.5.v1.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.5.v1.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.5.v1.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.5.w1.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.5.w1.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.5.w1.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.5.w2.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.5.w2.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.5.w2.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.6.v1.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.6.v1.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.6.v1.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.6.w1.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.6.w1.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.6.w1.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.6.w2.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.6.w2.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.6.w2.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.7.v1.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.7.v1.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.7.v1.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.7.w1.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.7.w1.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.7.w1.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.7.w2.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.7.w2.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.7.w2.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.8.v1.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.8.v1.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.8.v1.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.8.w1.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.8.w1.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.8.w1.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.8.w2.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.8.w2.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.8.w2.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.9.v1.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.9.v1.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.9.v1.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.9.w1.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.9.w1.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.9.w1.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.9.w2.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.9.w2.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.experts.9.w2.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.router.layer.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.router.layer.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.ffn.router.layer.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.norm_attn_norm.attn.Wqkv.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.norm_attn_norm.attn.Wqkv.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.norm_attn_norm.attn.Wqkv.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.norm_attn_norm.attn.out_proj.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.norm_attn_norm.attn.out_proj.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.norm_attn_norm.attn.out_proj.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.norm_attn_norm.norm_1.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.32.norm_attn_norm.norm_2.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.0.v1.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.0.v1.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.0.v1.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.0.w1.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.0.w1.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.0.w1.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.0.w2.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.0.w2.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.0.w2.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.1.v1.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.1.v1.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.1.v1.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.1.w1.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.1.w1.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.1.w1.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.1.w2.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.1.w2.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.1.w2.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.10.v1.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.10.v1.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.10.v1.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.10.w1.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.10.w1.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.10.w1.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.10.w2.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.10.w2.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.10.w2.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.11.v1.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.11.v1.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.11.v1.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.11.w1.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.11.w1.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.11.w1.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.11.w2.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.11.w2.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.11.w2.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.12.v1.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.12.v1.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.12.v1.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.12.w1.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.12.w1.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.12.w1.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.12.w2.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.12.w2.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.12.w2.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.13.v1.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.13.v1.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.13.v1.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.13.w1.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.13.w1.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.13.w1.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.13.w2.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.13.w2.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.13.w2.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.14.v1.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.14.v1.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.14.v1.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.14.w1.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.14.w1.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.14.w1.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.14.w2.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.14.w2.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.14.w2.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.15.v1.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.15.v1.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.15.v1.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.15.w1.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.15.w1.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.15.w1.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.15.w2.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.15.w2.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.15.w2.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.2.v1.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.2.v1.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.2.v1.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.2.w1.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.2.w1.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.2.w1.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.2.w2.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.2.w2.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.2.w2.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.3.v1.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.3.v1.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.3.v1.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.3.w1.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.3.w1.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.3.w1.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.3.w2.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.3.w2.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.3.w2.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.4.v1.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.4.v1.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.4.v1.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.4.w1.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.4.w1.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.4.w1.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.4.w2.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.4.w2.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.4.w2.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.5.v1.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.5.v1.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.5.v1.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.5.w1.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.5.w1.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.5.w1.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.5.w2.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.5.w2.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.5.w2.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.6.v1.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.6.v1.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.6.v1.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.6.w1.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.6.w1.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.6.w1.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.6.w2.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.6.w2.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.6.w2.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.7.v1.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.7.v1.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.7.v1.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.7.w1.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.7.w1.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.7.w1.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.7.w2.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.7.w2.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.7.w2.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.8.v1.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.8.v1.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.8.v1.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.8.w1.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.8.w1.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.8.w1.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.8.w2.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.8.w2.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.8.w2.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.9.v1.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.9.v1.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.9.v1.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.9.w1.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.9.w1.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.9.w1.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.9.w2.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.9.w2.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.experts.9.w2.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.router.layer.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.router.layer.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.ffn.router.layer.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.norm_attn_norm.attn.Wqkv.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.norm_attn_norm.attn.Wqkv.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.norm_attn_norm.attn.Wqkv.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.norm_attn_norm.attn.out_proj.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.norm_attn_norm.attn.out_proj.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.norm_attn_norm.attn.out_proj.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.norm_attn_norm.norm_1.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.33.norm_attn_norm.norm_2.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.0.v1.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.0.v1.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.0.v1.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.0.w1.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.0.w1.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.0.w1.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.0.w2.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.0.w2.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.0.w2.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.1.v1.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.1.v1.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.1.v1.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.1.w1.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.1.w1.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.1.w1.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.1.w2.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.1.w2.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.1.w2.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.10.v1.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.10.v1.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.10.v1.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.10.w1.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.10.w1.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.10.w1.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.10.w2.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.10.w2.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.10.w2.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.11.v1.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.11.v1.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.11.v1.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.11.w1.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.11.w1.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.11.w1.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.11.w2.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.11.w2.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.11.w2.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.12.v1.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.12.v1.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.12.v1.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.12.w1.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.12.w1.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.12.w1.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.12.w2.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.12.w2.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.12.w2.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.13.v1.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.13.v1.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.13.v1.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.13.w1.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.13.w1.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.13.w1.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.13.w2.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.13.w2.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.13.w2.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.14.v1.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.14.v1.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.14.v1.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.14.w1.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.14.w1.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.14.w1.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.14.w2.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.14.w2.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.14.w2.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.15.v1.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.15.v1.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.15.v1.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.15.w1.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.15.w1.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.15.w1.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.15.w2.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.15.w2.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.15.w2.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.2.v1.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.2.v1.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.2.v1.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.2.w1.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.2.w1.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.2.w1.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.2.w2.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.2.w2.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.2.w2.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.3.v1.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.3.v1.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.3.v1.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.3.w1.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.3.w1.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.3.w1.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.3.w2.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.3.w2.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.3.w2.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.4.v1.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.4.v1.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.4.v1.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.4.w1.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.4.w1.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.4.w1.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.4.w2.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.4.w2.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.4.w2.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.5.v1.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.5.v1.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.5.v1.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.5.w1.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.5.w1.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.5.w1.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.5.w2.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.5.w2.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.5.w2.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.6.v1.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.6.v1.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.6.v1.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.6.w1.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.6.w1.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.6.w1.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.6.w2.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.6.w2.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.6.w2.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.7.v1.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.7.v1.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.7.v1.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.7.w1.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.7.w1.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.7.w1.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.7.w2.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.7.w2.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.7.w2.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.8.v1.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.8.v1.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.8.v1.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.8.w1.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.8.w1.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.8.w1.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.8.w2.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.8.w2.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.8.w2.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.9.v1.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.9.v1.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.9.v1.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.9.w1.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.9.w1.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.9.w1.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.9.w2.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.9.w2.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.ffn.experts.9.w2.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.ffn.router.layer.biases": "model-00012-of-00014.safetensors",
+        "transformer.blocks.34.ffn.router.layer.scales": "model-00012-of-00014.safetensors",
+        "transformer.blocks.34.ffn.router.layer.weight": "model-00012-of-00014.safetensors",
+        "transformer.blocks.34.norm_attn_norm.attn.Wqkv.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.norm_attn_norm.attn.Wqkv.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.norm_attn_norm.attn.Wqkv.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.norm_attn_norm.attn.out_proj.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.norm_attn_norm.attn.out_proj.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.norm_attn_norm.attn.out_proj.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.norm_attn_norm.norm_1.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.34.norm_attn_norm.norm_2.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.0.v1.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.0.v1.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.0.v1.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.0.w1.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.0.w1.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.0.w1.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.0.w2.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.0.w2.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.0.w2.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.1.v1.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.1.v1.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.1.v1.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.1.w1.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.1.w1.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.1.w1.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.1.w2.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.1.w2.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.1.w2.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.10.v1.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.10.v1.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.10.v1.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.10.w1.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.10.w1.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.10.w1.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.10.w2.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.10.w2.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.10.w2.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.11.v1.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.11.v1.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.11.v1.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.11.w1.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.11.w1.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.11.w1.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.11.w2.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.11.w2.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.11.w2.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.12.v1.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.12.v1.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.12.v1.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.12.w1.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.12.w1.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.12.w1.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.12.w2.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.12.w2.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.12.w2.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.13.v1.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.13.v1.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.13.v1.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.13.w1.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.13.w1.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.13.w1.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.13.w2.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.13.w2.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.13.w2.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.14.v1.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.14.v1.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.14.v1.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.14.w1.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.14.w1.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.14.w1.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.14.w2.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.14.w2.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.14.w2.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.15.v1.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.15.v1.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.15.v1.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.15.w1.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.15.w1.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.15.w1.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.15.w2.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.15.w2.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.15.w2.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.2.v1.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.2.v1.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.2.v1.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.2.w1.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.2.w1.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.2.w1.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.2.w2.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.2.w2.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.2.w2.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.3.v1.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.3.v1.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.3.v1.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.3.w1.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.3.w1.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.3.w1.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.3.w2.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.3.w2.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.3.w2.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.4.v1.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.4.v1.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.4.v1.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.4.w1.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.4.w1.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.4.w1.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.4.w2.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.4.w2.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.4.w2.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.5.v1.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.5.v1.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.5.v1.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.5.w1.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.5.w1.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.5.w1.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.5.w2.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.5.w2.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.5.w2.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.6.v1.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.6.v1.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.6.v1.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.6.w1.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.6.w1.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.6.w1.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.6.w2.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.6.w2.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.6.w2.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.7.v1.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.7.v1.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.7.v1.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.7.w1.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.7.w1.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.7.w1.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.7.w2.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.7.w2.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.7.w2.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.8.v1.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.8.v1.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.8.v1.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.8.w1.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.8.w1.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.8.w1.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.8.w2.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.8.w2.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.8.w2.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.9.v1.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.9.v1.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.9.v1.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.9.w1.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.9.w1.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.9.w1.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.9.w2.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.9.w2.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.experts.9.w2.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.router.layer.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.router.layer.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.ffn.router.layer.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.norm_attn_norm.attn.Wqkv.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.norm_attn_norm.attn.Wqkv.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.norm_attn_norm.attn.Wqkv.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.norm_attn_norm.attn.out_proj.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.norm_attn_norm.attn.out_proj.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.norm_attn_norm.attn.out_proj.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.norm_attn_norm.norm_1.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.35.norm_attn_norm.norm_2.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.0.v1.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.0.v1.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.0.v1.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.0.w1.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.0.w1.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.0.w1.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.0.w2.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.0.w2.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.0.w2.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.1.v1.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.1.v1.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.1.v1.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.1.w1.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.1.w1.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.1.w1.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.1.w2.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.1.w2.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.1.w2.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.10.v1.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.10.v1.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.10.v1.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.10.w1.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.10.w1.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.10.w1.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.10.w2.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.10.w2.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.10.w2.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.11.v1.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.11.v1.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.11.v1.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.11.w1.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.11.w1.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.11.w1.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.11.w2.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.11.w2.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.11.w2.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.12.v1.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.12.v1.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.12.v1.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.12.w1.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.12.w1.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.12.w1.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.12.w2.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.12.w2.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.12.w2.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.13.v1.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.13.v1.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.13.v1.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.13.w1.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.13.w1.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.13.w1.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.13.w2.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.13.w2.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.13.w2.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.14.v1.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.14.v1.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.14.v1.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.14.w1.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.14.w1.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.14.w1.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.14.w2.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.14.w2.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.14.w2.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.15.v1.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.15.v1.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.15.v1.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.15.w1.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.15.w1.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.15.w1.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.15.w2.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.15.w2.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.15.w2.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.2.v1.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.2.v1.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.2.v1.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.2.w1.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.2.w1.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.2.w1.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.2.w2.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.2.w2.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.2.w2.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.3.v1.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.3.v1.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.3.v1.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.3.w1.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.3.w1.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.3.w1.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.3.w2.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.3.w2.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.3.w2.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.4.v1.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.4.v1.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.4.v1.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.4.w1.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.4.w1.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.4.w1.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.4.w2.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.4.w2.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.4.w2.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.5.v1.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.5.v1.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.5.v1.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.5.w1.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.5.w1.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.5.w1.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.5.w2.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.5.w2.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.5.w2.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.6.v1.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.6.v1.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.6.v1.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.6.w1.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.6.w1.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.6.w1.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.6.w2.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.6.w2.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.6.w2.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.7.v1.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.7.v1.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.7.v1.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.7.w1.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.7.w1.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.7.w1.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.7.w2.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.7.w2.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.7.w2.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.8.v1.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.8.v1.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.8.v1.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.8.w1.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.8.w1.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.8.w1.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.8.w2.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.8.w2.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.8.w2.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.9.v1.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.9.v1.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.9.v1.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.9.w1.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.9.w1.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.9.w1.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.9.w2.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.9.w2.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.experts.9.w2.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.router.layer.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.router.layer.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.ffn.router.layer.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.norm_attn_norm.attn.Wqkv.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.norm_attn_norm.attn.Wqkv.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.norm_attn_norm.attn.Wqkv.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.norm_attn_norm.attn.out_proj.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.norm_attn_norm.attn.out_proj.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.norm_attn_norm.attn.out_proj.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.norm_attn_norm.norm_1.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.36.norm_attn_norm.norm_2.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.0.v1.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.0.v1.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.0.v1.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.0.w1.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.0.w1.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.0.w1.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.0.w2.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.0.w2.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.0.w2.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.1.v1.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.1.v1.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.1.v1.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.1.w1.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.1.w1.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.1.w1.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.1.w2.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.1.w2.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.1.w2.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.10.v1.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.10.v1.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.10.v1.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.10.w1.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.10.w1.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.10.w1.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.10.w2.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.10.w2.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.10.w2.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.11.v1.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.11.v1.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.11.v1.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.11.w1.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.11.w1.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.11.w1.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.11.w2.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.11.w2.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.11.w2.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.12.v1.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.12.v1.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.12.v1.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.12.w1.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.12.w1.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.12.w1.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.12.w2.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.12.w2.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.12.w2.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.13.v1.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.13.v1.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.13.v1.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.13.w1.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.13.w1.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.13.w1.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.13.w2.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.13.w2.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.13.w2.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.14.v1.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.14.v1.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.14.v1.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.14.w1.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.14.w1.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.14.w1.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.14.w2.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.14.w2.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.14.w2.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.15.v1.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.15.v1.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.15.v1.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.15.w1.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.15.w1.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.15.w1.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.15.w2.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.15.w2.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.15.w2.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.2.v1.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.2.v1.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.2.v1.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.2.w1.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.2.w1.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.2.w1.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.2.w2.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.2.w2.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.2.w2.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.3.v1.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.3.v1.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.3.v1.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.3.w1.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.3.w1.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.3.w1.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.3.w2.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.3.w2.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.3.w2.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.4.v1.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.4.v1.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.4.v1.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.4.w1.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.4.w1.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.4.w1.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.4.w2.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.4.w2.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.4.w2.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.5.v1.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.5.v1.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.5.v1.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.5.w1.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.5.w1.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.5.w1.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.5.w2.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.5.w2.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.5.w2.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.6.v1.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.6.v1.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.6.v1.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.6.w1.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.6.w1.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.6.w1.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.6.w2.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.6.w2.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.6.w2.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.7.v1.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.7.v1.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.7.v1.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.7.w1.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.7.w1.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.7.w1.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.7.w2.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.7.w2.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.7.w2.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.8.v1.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.8.v1.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.8.v1.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.8.w1.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.8.w1.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.8.w1.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.8.w2.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.8.w2.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.8.w2.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.9.v1.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.9.v1.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.9.v1.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.9.w1.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.9.w1.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.9.w1.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.9.w2.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.9.w2.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.experts.9.w2.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.ffn.router.layer.biases": "model-00013-of-00014.safetensors",
+        "transformer.blocks.37.ffn.router.layer.scales": "model-00013-of-00014.safetensors",
+        "transformer.blocks.37.ffn.router.layer.weight": "model-00013-of-00014.safetensors",
+        "transformer.blocks.37.norm_attn_norm.attn.Wqkv.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.norm_attn_norm.attn.Wqkv.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.norm_attn_norm.attn.Wqkv.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.norm_attn_norm.attn.out_proj.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.norm_attn_norm.attn.out_proj.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.norm_attn_norm.attn.out_proj.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.norm_attn_norm.norm_1.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.37.norm_attn_norm.norm_2.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.0.v1.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.0.v1.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.0.v1.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.0.w1.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.0.w1.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.0.w1.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.0.w2.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.0.w2.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.0.w2.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.1.v1.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.1.v1.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.1.v1.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.1.w1.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.1.w1.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.1.w1.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.1.w2.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.1.w2.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.1.w2.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.10.v1.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.10.v1.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.10.v1.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.10.w1.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.10.w1.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.10.w1.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.10.w2.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.10.w2.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.10.w2.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.11.v1.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.11.v1.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.11.v1.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.11.w1.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.11.w1.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.11.w1.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.11.w2.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.11.w2.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.11.w2.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.12.v1.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.12.v1.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.12.v1.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.12.w1.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.12.w1.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.12.w1.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.12.w2.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.12.w2.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.12.w2.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.13.v1.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.13.v1.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.13.v1.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.13.w1.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.13.w1.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.13.w1.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.13.w2.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.13.w2.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.13.w2.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.14.v1.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.14.v1.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.14.v1.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.14.w1.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.14.w1.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.14.w1.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.14.w2.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.14.w2.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.14.w2.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.15.v1.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.15.v1.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.15.v1.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.15.w1.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.15.w1.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.15.w1.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.15.w2.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.15.w2.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.15.w2.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.2.v1.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.2.v1.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.2.v1.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.2.w1.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.2.w1.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.2.w1.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.2.w2.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.2.w2.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.2.w2.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.3.v1.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.3.v1.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.3.v1.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.3.w1.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.3.w1.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.3.w1.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.3.w2.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.3.w2.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.3.w2.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.4.v1.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.4.v1.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.4.v1.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.4.w1.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.4.w1.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.4.w1.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.4.w2.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.4.w2.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.4.w2.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.5.v1.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.5.v1.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.5.v1.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.5.w1.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.5.w1.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.5.w1.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.5.w2.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.5.w2.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.5.w2.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.6.v1.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.6.v1.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.6.v1.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.6.w1.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.6.w1.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.6.w1.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.6.w2.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.6.w2.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.6.w2.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.7.v1.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.7.v1.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.7.v1.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.7.w1.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.7.w1.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.7.w1.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.7.w2.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.7.w2.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.7.w2.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.8.v1.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.8.v1.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.8.v1.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.8.w1.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.8.w1.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.8.w1.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.8.w2.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.8.w2.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.8.w2.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.9.v1.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.9.v1.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.9.v1.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.9.w1.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.9.w1.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.9.w1.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.9.w2.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.9.w2.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.experts.9.w2.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.router.layer.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.router.layer.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.ffn.router.layer.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.norm_attn_norm.attn.Wqkv.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.norm_attn_norm.attn.Wqkv.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.norm_attn_norm.attn.Wqkv.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.norm_attn_norm.attn.out_proj.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.norm_attn_norm.attn.out_proj.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.norm_attn_norm.attn.out_proj.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.norm_attn_norm.norm_1.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.38.norm_attn_norm.norm_2.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.0.v1.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.0.v1.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.0.v1.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.0.w1.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.0.w1.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.0.w1.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.0.w2.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.0.w2.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.0.w2.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.1.v1.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.1.v1.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.1.v1.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.1.w1.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.1.w1.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.1.w1.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.1.w2.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.1.w2.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.1.w2.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.10.v1.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.10.v1.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.10.v1.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.10.w1.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.10.w1.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.10.w1.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.10.w2.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.10.w2.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.10.w2.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.11.v1.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.11.v1.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.11.v1.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.11.w1.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.11.w1.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.11.w1.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.11.w2.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.11.w2.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.11.w2.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.12.v1.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.12.v1.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.12.v1.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.12.w1.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.12.w1.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.12.w1.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.12.w2.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.12.w2.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.12.w2.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.13.v1.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.13.v1.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.13.v1.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.13.w1.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.13.w1.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.13.w1.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.13.w2.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.13.w2.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.13.w2.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.14.v1.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.14.v1.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.14.v1.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.14.w1.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.14.w1.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.14.w1.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.14.w2.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.14.w2.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.14.w2.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.15.v1.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.15.v1.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.15.v1.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.15.w1.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.15.w1.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.15.w1.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.15.w2.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.15.w2.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.15.w2.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.2.v1.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.2.v1.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.2.v1.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.2.w1.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.2.w1.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.2.w1.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.2.w2.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.2.w2.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.2.w2.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.3.v1.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.3.v1.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.3.v1.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.3.w1.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.3.w1.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.3.w1.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.3.w2.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.3.w2.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.3.w2.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.4.v1.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.4.v1.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.4.v1.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.4.w1.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.4.w1.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.4.w1.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.4.w2.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.4.w2.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.4.w2.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.5.v1.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.5.v1.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.5.v1.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.5.w1.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.5.w1.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.5.w1.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.5.w2.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.5.w2.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.5.w2.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.6.v1.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.6.v1.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.6.v1.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.6.w1.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.6.w1.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.6.w1.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.6.w2.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.6.w2.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.6.w2.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.7.v1.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.7.v1.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.7.v1.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.7.w1.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.7.w1.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.7.w1.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.7.w2.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.7.w2.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.7.w2.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.8.v1.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.8.v1.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.8.v1.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.8.w1.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.8.w1.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.8.w1.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.8.w2.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.8.w2.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.8.w2.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.9.v1.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.9.v1.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.9.v1.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.9.w1.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.9.w1.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.9.w1.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.9.w2.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.9.w2.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.experts.9.w2.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.router.layer.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.router.layer.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.ffn.router.layer.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.norm_attn_norm.attn.Wqkv.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.norm_attn_norm.attn.Wqkv.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.norm_attn_norm.attn.Wqkv.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.norm_attn_norm.attn.out_proj.biases": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.norm_attn_norm.attn.out_proj.scales": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.norm_attn_norm.attn.out_proj.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.norm_attn_norm.norm_1.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.39.norm_attn_norm.norm_2.weight": "model-00014-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.0.v1.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.0.v1.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.0.v1.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.0.w1.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.0.w1.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.0.w1.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.0.w2.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.0.w2.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.0.w2.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.1.v1.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.1.v1.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.1.v1.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.1.w1.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.1.w1.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.1.w1.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.1.w2.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.1.w2.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.1.w2.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.10.v1.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.10.v1.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.10.v1.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.10.w1.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.10.w1.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.10.w1.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.10.w2.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.10.w2.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.10.w2.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.11.v1.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.11.v1.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.11.v1.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.11.w1.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.11.w1.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.11.w1.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.11.w2.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.11.w2.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.11.w2.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.12.v1.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.12.v1.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.12.v1.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.12.w1.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.12.w1.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.12.w1.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.12.w2.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.12.w2.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.12.w2.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.13.v1.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.13.v1.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.13.v1.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.13.w1.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.13.w1.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.13.w1.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.13.w2.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.13.w2.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.13.w2.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.14.v1.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.14.v1.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.14.v1.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.14.w1.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.14.w1.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.14.w1.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.14.w2.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.14.w2.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.14.w2.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.15.v1.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.15.v1.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.15.v1.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.15.w1.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.15.w1.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.15.w1.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.15.w2.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.15.w2.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.15.w2.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.2.v1.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.2.v1.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.2.v1.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.2.w1.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.2.w1.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.2.w1.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.2.w2.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.2.w2.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.2.w2.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.3.v1.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.3.v1.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.3.v1.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.3.w1.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.3.w1.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.3.w1.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.3.w2.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.3.w2.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.3.w2.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.4.v1.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.4.v1.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.4.v1.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.4.w1.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.4.w1.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.4.w1.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.4.w2.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.4.w2.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.4.w2.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.5.v1.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.5.v1.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.5.v1.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.5.w1.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.5.w1.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.5.w1.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.5.w2.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.5.w2.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.5.w2.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.6.v1.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.6.v1.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.6.v1.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.6.w1.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.6.w1.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.6.w1.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.6.w2.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.6.w2.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.6.w2.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.7.v1.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.7.v1.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.7.v1.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.7.w1.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.7.w1.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.7.w1.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.7.w2.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.7.w2.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.7.w2.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.8.v1.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.8.v1.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.8.v1.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.8.w1.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.8.w1.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.8.w1.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.8.w2.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.8.w2.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.8.w2.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.9.v1.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.9.v1.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.9.v1.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.9.w1.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.9.w1.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.9.w1.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.9.w2.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.9.w2.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.experts.9.w2.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.router.layer.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.router.layer.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.ffn.router.layer.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.norm_attn_norm.attn.Wqkv.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.norm_attn_norm.attn.Wqkv.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.norm_attn_norm.attn.Wqkv.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.norm_attn_norm.attn.out_proj.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.norm_attn_norm.attn.out_proj.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.norm_attn_norm.attn.out_proj.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.norm_attn_norm.norm_1.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.4.norm_attn_norm.norm_2.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.0.v1.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.0.v1.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.0.v1.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.0.w1.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.0.w1.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.0.w1.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.0.w2.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.0.w2.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.0.w2.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.1.v1.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.1.v1.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.1.v1.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.1.w1.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.1.w1.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.1.w1.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.1.w2.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.1.w2.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.1.w2.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.10.v1.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.10.v1.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.10.v1.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.10.w1.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.10.w1.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.10.w1.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.10.w2.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.10.w2.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.10.w2.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.11.v1.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.11.v1.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.11.v1.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.11.w1.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.11.w1.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.11.w1.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.11.w2.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.11.w2.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.11.w2.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.12.v1.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.12.v1.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.12.v1.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.12.w1.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.12.w1.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.12.w1.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.12.w2.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.12.w2.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.12.w2.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.13.v1.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.13.v1.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.13.v1.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.13.w1.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.13.w1.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.13.w1.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.13.w2.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.13.w2.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.13.w2.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.14.v1.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.14.v1.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.14.v1.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.14.w1.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.14.w1.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.14.w1.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.14.w2.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.14.w2.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.14.w2.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.15.v1.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.15.v1.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.15.v1.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.15.w1.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.15.w1.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.15.w1.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.15.w2.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.15.w2.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.15.w2.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.2.v1.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.2.v1.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.2.v1.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.2.w1.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.2.w1.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.2.w1.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.2.w2.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.2.w2.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.2.w2.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.3.v1.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.3.v1.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.3.v1.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.3.w1.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.3.w1.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.3.w1.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.3.w2.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.3.w2.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.3.w2.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.4.v1.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.4.v1.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.4.v1.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.4.w1.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.4.w1.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.4.w1.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.4.w2.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.4.w2.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.4.w2.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.5.v1.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.5.v1.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.5.v1.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.5.w1.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.5.w1.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.5.w1.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.5.w2.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.5.w2.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.5.w2.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.6.v1.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.6.v1.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.6.v1.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.6.w1.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.6.w1.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.6.w1.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.6.w2.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.6.w2.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.6.w2.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.7.v1.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.7.v1.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.7.v1.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.7.w1.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.7.w1.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.7.w1.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.7.w2.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.7.w2.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.7.w2.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.8.v1.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.8.v1.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.8.v1.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.8.w1.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.8.w1.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.8.w1.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.8.w2.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.8.w2.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.8.w2.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.9.v1.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.9.v1.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.9.v1.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.9.w1.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.9.w1.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.9.w1.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.9.w2.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.9.w2.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.experts.9.w2.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.ffn.router.layer.biases": "model-00002-of-00014.safetensors",
+        "transformer.blocks.5.ffn.router.layer.scales": "model-00002-of-00014.safetensors",
+        "transformer.blocks.5.ffn.router.layer.weight": "model-00002-of-00014.safetensors",
+        "transformer.blocks.5.norm_attn_norm.attn.Wqkv.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.norm_attn_norm.attn.Wqkv.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.norm_attn_norm.attn.Wqkv.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.norm_attn_norm.attn.out_proj.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.norm_attn_norm.attn.out_proj.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.norm_attn_norm.attn.out_proj.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.norm_attn_norm.norm_1.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.5.norm_attn_norm.norm_2.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.0.v1.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.0.v1.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.0.v1.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.0.w1.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.0.w1.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.0.w1.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.0.w2.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.0.w2.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.0.w2.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.1.v1.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.1.v1.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.1.v1.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.1.w1.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.1.w1.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.1.w1.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.1.w2.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.1.w2.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.1.w2.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.10.v1.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.10.v1.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.10.v1.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.10.w1.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.10.w1.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.10.w1.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.10.w2.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.10.w2.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.10.w2.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.11.v1.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.11.v1.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.11.v1.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.11.w1.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.11.w1.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.11.w1.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.11.w2.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.11.w2.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.11.w2.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.12.v1.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.12.v1.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.12.v1.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.12.w1.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.12.w1.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.12.w1.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.12.w2.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.12.w2.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.12.w2.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.13.v1.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.13.v1.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.13.v1.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.13.w1.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.13.w1.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.13.w1.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.13.w2.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.13.w2.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.13.w2.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.14.v1.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.14.v1.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.14.v1.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.14.w1.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.14.w1.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.14.w1.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.14.w2.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.14.w2.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.14.w2.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.15.v1.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.15.v1.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.15.v1.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.15.w1.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.15.w1.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.15.w1.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.15.w2.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.15.w2.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.15.w2.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.2.v1.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.2.v1.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.2.v1.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.2.w1.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.2.w1.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.2.w1.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.2.w2.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.2.w2.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.2.w2.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.3.v1.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.3.v1.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.3.v1.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.3.w1.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.3.w1.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.3.w1.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.3.w2.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.3.w2.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.3.w2.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.4.v1.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.4.v1.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.4.v1.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.4.w1.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.4.w1.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.4.w1.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.4.w2.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.4.w2.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.4.w2.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.5.v1.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.5.v1.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.5.v1.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.5.w1.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.5.w1.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.5.w1.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.5.w2.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.5.w2.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.5.w2.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.6.v1.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.6.v1.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.6.v1.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.6.w1.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.6.w1.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.6.w1.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.6.w2.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.6.w2.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.6.w2.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.7.v1.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.7.v1.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.7.v1.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.7.w1.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.7.w1.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.7.w1.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.7.w2.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.7.w2.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.7.w2.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.8.v1.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.8.v1.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.8.v1.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.8.w1.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.8.w1.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.8.w1.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.8.w2.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.8.w2.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.8.w2.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.9.v1.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.9.v1.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.9.v1.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.9.w1.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.9.w1.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.9.w1.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.9.w2.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.9.w2.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.experts.9.w2.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.router.layer.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.router.layer.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.ffn.router.layer.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.norm_attn_norm.attn.Wqkv.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.norm_attn_norm.attn.Wqkv.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.norm_attn_norm.attn.Wqkv.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.norm_attn_norm.attn.out_proj.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.norm_attn_norm.attn.out_proj.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.norm_attn_norm.attn.out_proj.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.norm_attn_norm.norm_1.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.6.norm_attn_norm.norm_2.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.0.v1.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.0.v1.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.0.v1.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.0.w1.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.0.w1.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.0.w1.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.0.w2.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.0.w2.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.0.w2.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.1.v1.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.1.v1.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.1.v1.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.1.w1.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.1.w1.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.1.w1.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.1.w2.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.1.w2.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.1.w2.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.10.v1.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.10.v1.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.10.v1.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.10.w1.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.10.w1.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.10.w1.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.10.w2.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.10.w2.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.10.w2.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.11.v1.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.11.v1.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.11.v1.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.11.w1.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.11.w1.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.11.w1.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.11.w2.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.11.w2.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.11.w2.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.12.v1.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.12.v1.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.12.v1.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.12.w1.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.12.w1.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.12.w1.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.12.w2.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.12.w2.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.12.w2.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.13.v1.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.13.v1.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.13.v1.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.13.w1.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.13.w1.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.13.w1.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.13.w2.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.13.w2.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.13.w2.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.14.v1.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.14.v1.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.14.v1.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.14.w1.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.14.w1.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.14.w1.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.14.w2.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.14.w2.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.14.w2.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.15.v1.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.15.v1.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.15.v1.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.15.w1.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.15.w1.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.15.w1.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.15.w2.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.15.w2.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.15.w2.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.2.v1.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.2.v1.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.2.v1.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.2.w1.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.2.w1.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.2.w1.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.2.w2.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.2.w2.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.2.w2.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.3.v1.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.3.v1.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.3.v1.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.3.w1.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.3.w1.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.3.w1.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.3.w2.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.3.w2.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.3.w2.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.4.v1.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.4.v1.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.4.v1.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.4.w1.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.4.w1.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.4.w1.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.4.w2.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.4.w2.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.4.w2.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.5.v1.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.5.v1.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.5.v1.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.5.w1.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.5.w1.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.5.w1.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.5.w2.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.5.w2.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.5.w2.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.6.v1.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.6.v1.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.6.v1.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.6.w1.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.6.w1.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.6.w1.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.6.w2.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.6.w2.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.6.w2.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.7.v1.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.7.v1.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.7.v1.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.7.w1.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.7.w1.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.7.w1.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.7.w2.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.7.w2.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.7.w2.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.8.v1.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.8.v1.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.8.v1.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.8.w1.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.8.w1.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.8.w1.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.8.w2.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.8.w2.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.8.w2.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.9.v1.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.9.v1.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.9.v1.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.9.w1.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.9.w1.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.9.w1.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.9.w2.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.9.w2.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.experts.9.w2.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.router.layer.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.router.layer.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.ffn.router.layer.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.norm_attn_norm.attn.Wqkv.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.norm_attn_norm.attn.Wqkv.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.norm_attn_norm.attn.Wqkv.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.norm_attn_norm.attn.out_proj.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.norm_attn_norm.attn.out_proj.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.norm_attn_norm.attn.out_proj.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.norm_attn_norm.norm_1.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.7.norm_attn_norm.norm_2.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.0.v1.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.0.v1.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.0.v1.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.0.w1.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.0.w1.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.0.w1.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.0.w2.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.0.w2.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.0.w2.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.1.v1.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.1.v1.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.1.v1.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.1.w1.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.1.w1.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.1.w1.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.1.w2.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.1.w2.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.1.w2.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.10.v1.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.10.v1.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.10.v1.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.10.w1.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.10.w1.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.10.w1.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.10.w2.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.10.w2.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.10.w2.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.11.v1.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.11.v1.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.11.v1.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.11.w1.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.11.w1.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.11.w1.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.11.w2.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.11.w2.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.11.w2.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.12.v1.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.12.v1.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.12.v1.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.12.w1.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.12.w1.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.12.w1.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.12.w2.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.12.w2.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.12.w2.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.13.v1.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.13.v1.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.13.v1.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.13.w1.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.13.w1.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.13.w1.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.13.w2.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.13.w2.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.13.w2.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.14.v1.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.14.v1.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.14.v1.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.14.w1.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.14.w1.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.14.w1.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.14.w2.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.14.w2.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.14.w2.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.15.v1.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.15.v1.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.15.v1.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.15.w1.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.15.w1.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.15.w1.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.15.w2.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.15.w2.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.15.w2.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.2.v1.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.2.v1.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.2.v1.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.2.w1.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.2.w1.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.2.w1.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.2.w2.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.2.w2.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.2.w2.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.3.v1.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.3.v1.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.3.v1.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.3.w1.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.3.w1.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.3.w1.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.3.w2.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.3.w2.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.3.w2.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.4.v1.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.4.v1.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.4.v1.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.4.w1.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.4.w1.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.4.w1.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.4.w2.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.4.w2.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.4.w2.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.5.v1.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.5.v1.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.5.v1.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.5.w1.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.5.w1.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.5.w1.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.5.w2.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.5.w2.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.5.w2.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.6.v1.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.6.v1.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.6.v1.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.6.w1.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.6.w1.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.6.w1.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.6.w2.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.6.w2.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.6.w2.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.7.v1.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.7.v1.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.7.v1.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.7.w1.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.7.w1.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.7.w1.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.7.w2.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.7.w2.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.7.w2.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.8.v1.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.8.v1.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.8.v1.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.8.w1.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.8.w1.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.8.w1.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.8.w2.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.8.w2.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.8.w2.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.9.v1.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.9.v1.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.9.v1.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.9.w1.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.9.w1.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.9.w1.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.9.w2.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.9.w2.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.experts.9.w2.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.ffn.router.layer.biases": "model-00003-of-00014.safetensors",
+        "transformer.blocks.8.ffn.router.layer.scales": "model-00003-of-00014.safetensors",
+        "transformer.blocks.8.ffn.router.layer.weight": "model-00003-of-00014.safetensors",
+        "transformer.blocks.8.norm_attn_norm.attn.Wqkv.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.norm_attn_norm.attn.Wqkv.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.norm_attn_norm.attn.Wqkv.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.norm_attn_norm.attn.out_proj.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.norm_attn_norm.attn.out_proj.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.norm_attn_norm.attn.out_proj.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.norm_attn_norm.norm_1.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.8.norm_attn_norm.norm_2.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.0.v1.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.0.v1.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.0.v1.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.0.w1.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.0.w1.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.0.w1.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.0.w2.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.0.w2.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.0.w2.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.1.v1.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.1.v1.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.1.v1.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.1.w1.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.1.w1.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.1.w1.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.1.w2.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.1.w2.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.1.w2.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.10.v1.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.10.v1.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.10.v1.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.10.w1.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.10.w1.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.10.w1.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.10.w2.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.10.w2.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.10.w2.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.11.v1.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.11.v1.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.11.v1.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.11.w1.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.11.w1.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.11.w1.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.11.w2.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.11.w2.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.11.w2.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.12.v1.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.12.v1.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.12.v1.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.12.w1.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.12.w1.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.12.w1.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.12.w2.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.12.w2.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.12.w2.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.13.v1.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.13.v1.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.13.v1.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.13.w1.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.13.w1.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.13.w1.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.13.w2.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.13.w2.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.13.w2.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.14.v1.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.14.v1.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.14.v1.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.14.w1.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.14.w1.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.14.w1.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.14.w2.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.14.w2.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.14.w2.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.15.v1.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.15.v1.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.15.v1.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.15.w1.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.15.w1.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.15.w1.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.15.w2.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.15.w2.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.15.w2.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.2.v1.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.2.v1.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.2.v1.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.2.w1.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.2.w1.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.2.w1.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.2.w2.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.2.w2.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.2.w2.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.3.v1.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.3.v1.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.3.v1.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.3.w1.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.3.w1.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.3.w1.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.3.w2.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.3.w2.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.3.w2.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.4.v1.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.4.v1.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.4.v1.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.4.w1.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.4.w1.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.4.w1.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.4.w2.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.4.w2.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.4.w2.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.5.v1.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.5.v1.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.5.v1.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.5.w1.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.5.w1.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.5.w1.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.5.w2.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.5.w2.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.5.w2.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.6.v1.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.6.v1.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.6.v1.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.6.w1.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.6.w1.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.6.w1.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.6.w2.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.6.w2.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.6.w2.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.7.v1.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.7.v1.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.7.v1.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.7.w1.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.7.w1.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.7.w1.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.7.w2.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.7.w2.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.7.w2.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.8.v1.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.8.v1.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.8.v1.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.8.w1.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.8.w1.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.8.w1.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.8.w2.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.8.w2.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.8.w2.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.9.v1.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.9.v1.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.9.v1.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.9.w1.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.9.w1.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.9.w1.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.9.w2.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.9.w2.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.experts.9.w2.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.router.layer.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.router.layer.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.ffn.router.layer.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.norm_attn_norm.attn.Wqkv.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.norm_attn_norm.attn.Wqkv.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.norm_attn_norm.attn.Wqkv.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.norm_attn_norm.attn.out_proj.biases": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.norm_attn_norm.attn.out_proj.scales": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.norm_attn_norm.attn.out_proj.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.norm_attn_norm.norm_1.weight": "model-00004-of-00014.safetensors",
+        "transformer.blocks.9.norm_attn_norm.norm_2.weight": "model-00004-of-00014.safetensors",
+        "transformer.norm_f.weight": "model-00014-of-00014.safetensors",
+        "transformer.wte.weight": "model-00001-of-00014.safetensors"
+    }
+}
\ No newline at end of file