Upload folder using huggingface_hub

README.md

@@ -65,7 +65,7 @@ Baichuan-M2 采用了三个核心技术创新：首先通过**大型验证器系
 
 ### 医疗领域适应
 - **Mid-Training**：医疗知识注入的同时保持通用能力
-- **强化学习**：多阶段 RL 策略优化
+- **端到端强化学习**：多阶段 RL 策略优化
 - **通专兼顾**：2:2:1 配比的医疗、通用、数学数据
 
 ## 🔧 快速开始

config.json

@@ -8,7 +8,7 @@
   "hidden_size": 5120,
   "initializer_range": 0.02,
   "intermediate_size": 27648,
-  "max_position_embeddings": 132768,
+  "max_position_embeddings": 131072,
   "max_window_layers": 64,
   "model_type": "qwen2",
   "num_attention_heads": 40,

draft/config.json

@@ -0,0 +1,29 @@
+{
+  "architectures": [
+    "LlamaForCausalLMEagle3"
+  ],
+  "attention_dropout": 0.0,
+  "eos_token_id": 151645,
+  "hidden_act": "silu",
+  "hidden_size": 5120,
+  "initializer_range": 0.02,
+  "intermediate_size": 27648,
+  "max_position_embeddings": 32768,
+  "max_window_layers": 64,
+  "model_type": "llama",
+  "num_attention_heads": 40,
+  "num_hidden_layers": 1,
+  "num_key_value_heads": 8,
+  "pad_token_id": 151643,
+  "rms_norm_eps": 1e-05,
+  "rope_scaling": null,
+  "rope_theta": 1000000.0,
+  "sliding_window": 131072,
+  "tie_word_embeddings": false,
+  "torch_dtype": "bfloat16",
+  "transformers_version": "4.51.3",
+  "use_cache": true,
+  "use_sliding_window": false,
+  "vocab_size": 152064,
+  "draft_vocab_size": 32000 
+}

draft/pytorch_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8cd48c028f13ee92fe60b99509fb6e8630ca87a1d48150a56ebe6dbd1b12f0bc
+size 1534046974

tokenizer_config.json

@@ -232,7 +232,7 @@
   "eos_token": "<|im_end|>",
   "errors": "replace",
   "extra_special_tokens": {},
-  "model_max_length": 132768,
+  "model_max_length": 131072,
   "pad_token": "<|endoftext|>",
   "split_special_tokens": false,
   "tokenizer_class": "Qwen2Tokenizer",