Synchronizing local compiler cache.

.gitattributes

@@ -5683,3 +5683,5 @@ neuronxcc-2.15.128.0+56dc5a86/MODULE_7abcc1f4fc8a5be9a24e+39f12043/model.neff fi
 neuronxcc-2.15.128.0+56dc5a86/MODULE_eadf8f3cb57245e99bf7+39f12043/model.neff filter=lfs diff=lfs merge=lfs -text
 neuronxcc-2.15.128.0+56dc5a86/MODULE_97bc2ac4ae7c07475ae9+39f12043/model.neff filter=lfs diff=lfs merge=lfs -text
 neuronxcc-2.15.128.0+56dc5a86/MODULE_9ac132d9fc76d1c18a07+39f12043/model.neff filter=lfs diff=lfs merge=lfs -text
+neuronxcc-2.15.128.0+56dc5a86/MODULE_008c4346863d39436b21+39f12043/model.neff filter=lfs diff=lfs merge=lfs -text
+neuronxcc-2.15.128.0+56dc5a86/MODULE_c76967a3d431b4607b48+39f12043/model.neff filter=lfs diff=lfs merge=lfs -text

neuronxcc-2.15.128.0+56dc5a86/0_REGISTRY/0.0.25/inference/llama/NousResearch/Meta-Llama-3-8B-Instruct/dc81de08276908ae6c69.json

@@ -0,0 +1 @@
+{"architectures": ["LlamaForCausalLM"], "attention_bias": false, "attention_dropout": 0.0, "bos_token_id": 128000, "eos_token_id": 128009, "hidden_act": "silu", "hidden_size": 4096, "initializer_range": 0.02, "intermediate_size": 14336, "max_position_embeddings": 8192, "mlp_bias": false, "model_type": "llama", "neuron": {"auto_cast_type": "bf16", "batch_size": 32, "checkpoint_id": "NousResearch/Meta-Llama-3-8B-Instruct", "checkpoint_revision": "53346005fb0ef11d3b6a83b12c895cca40156b6c", "compiler_type": "neuronx-cc", "compiler_version": "2.15.128.0+56dc5a86", "num_cores": 8, "sequence_length": 4096, "task": "text-generation"}, "num_attention_heads": 32, "num_hidden_layers": 32, "num_key_value_heads": 8, "pretraining_tp": 1, "rms_norm_eps": 1e-05, "rope_scaling": null, "rope_theta": 500000.0, "tie_word_embeddings": false, "torch_dtype": "bfloat16", "use_cache": true, "vocab_size": 128256}

neuronxcc-2.15.128.0+56dc5a86/MODULE_008c4346863d39436b21+39f12043/compile_flags.json

@@ -0,0 +1 @@
+["--model-type=transformer", "--auto-cast=none", "--execute-repetition=1"]

neuronxcc-2.15.128.0+56dc5a86/MODULE_008c4346863d39436b21+39f12043/model.hlo_module.pb

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:34acff9f650171fda4e73e8a6dee0b4b20b3e2c5d8ba3c55a774f0075555f2ad
+size 449021

neuronxcc-2.15.128.0+56dc5a86/MODULE_008c4346863d39436b21+39f12043/model.neff

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fbe9f8836b0d45f2f555ec7ab9b29c712faaf6b2613b54e36047d97ea77fab9b
+size 30188544

neuronxcc-2.15.128.0+56dc5a86/MODULE_c76967a3d431b4607b48+39f12043/compile_flags.json

@@ -0,0 +1 @@
+["--model-type=transformer", "--auto-cast=none", "--execute-repetition=1"]

neuronxcc-2.15.128.0+56dc5a86/MODULE_c76967a3d431b4607b48+39f12043/model.hlo_module.pb

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c073e293fbdf37af18003cef223472eb5a07cd2b0f71119d897f112e32c092df
+size 363555

neuronxcc-2.15.128.0+56dc5a86/MODULE_c76967a3d431b4607b48+39f12043/model.neff

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ac55681efd502ec146f4dadc1b39339b4e282e20f967a4450257edec32a33930
+size 5735424