Upload 3 files

README.md

@@ -0,0 +1,19 @@
+---
+inference: false
+language:
+- en
+library_name: transformers
+pipeline_tag: text-generation
+tags:
+- mixtral
+license: apache-2.0
+datasets:
+- jondurbin/airoboros-3.2
+---
+
+# Air-Striker-Mixtral-8x7B-ZLoss
+
+Experimental model, trained using config and [Transformers/Axolotl](https://github.com/DocShotgun/axolotl) forks provided by [Doctor-Shotgun](https://huggingface.co/Doctor-Shotgun)
+
+Model was fine-tuned from [Mixtral-8x7B-v0.1](https://huggingface.co/mistralai/Mixtral-8x7B-v0.1) with airoboros-3.2 dataset, for 4 epochs, ChatML prompt format at 8K context length.
+

adapter_config.json

@@ -0,0 +1,32 @@
+{
+  "alpha_pattern": {},
+  "auto_mapping": null,
+  "base_model_name_or_path": "/users/ubuntu/models/Mixtral-8x7B-v0.1",
+  "bias": "none",
+  "fan_in_fan_out": null,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 16,
+  "lora_dropout": 0.07,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "q_proj",
+    "w1",
+    "v_proj",
+    "gate",
+    "o_proj",
+    "w3",
+    "k_proj",
+    "w2"
+  ],
+  "task_type": "CAUSAL_LM"
+}

config.json

@@ -1,5 +1,5 @@
 {
-  "_name_or_path": "/home/hien/models/Mixtral-8x7B-v0.1",
+  "_name_or_path": "/users/ubuntu/models/Mixtral-8x7B-v0.1",
   "architectures": [
     "MixtralForCausalLM"
   ],
@@ -17,14 +17,27 @@
   "num_hidden_layers": 32,
   "num_key_value_heads": 8,
   "num_local_experts": 8,
-  "output_router_logits": false,
+  "output_router_logits": true,
+  "quantization_config": {
+    "bnb_4bit_compute_dtype": "bfloat16",
+    "bnb_4bit_quant_type": "nf4",
+    "bnb_4bit_use_double_quant": true,
+    "llm_int8_enable_fp32_cpu_offload": false,
+    "llm_int8_has_fp16_weight": false,
+    "llm_int8_skip_modules": null,
+    "llm_int8_threshold": 6.0,
+    "load_in_4bit": true,
+    "load_in_8bit": false,
+    "quant_method": "bitsandbytes"
+  },
   "rms_norm_eps": 1e-05,
   "rope_theta": 1000000.0,
   "router_aux_loss_coef": 0.02,
+  "router_z_loss_coef": 0.001,
   "sliding_window": null,
   "tie_word_embeddings": false,
-  "torch_dtype": "float16",
+  "torch_dtype": "bfloat16",
   "transformers_version": "4.37.0.dev0",
-  "use_cache": true,
+  "use_cache": false,
   "vocab_size": 32000
 }