Add files using upload-large-folder tool

.ipynb_checkpoints/README-checkpoint.md

@@ -0,0 +1,79 @@
+---
+license: apache-2.0
+tags:
+  - 中医大模型
+
+#model-type:
+##如 gpt、phi、llama、chatglm、baichuan 等
+#- gpt
+
+#domain:
+##如 nlp、cv、audio、multi-modal
+#- nlp
+
+#language:
+##语言代码列表 https://help.aliyun.com/document_detail/215387.html?spm=a2c4g.11186623.0.0.9f8d7467kni6Aa
+#- cn
+
+#metrics:
+##如 CIDEr、Blue、ROUGE 等
+#- CIDEr
+
+#tags:
+##各种自定义，包括 pretrained、fine-tuned、instruction-tuned、RL-tuned 等训练方法和其他
+#- pretrained
+
+#tools:
+##如 vllm、fastchat、llamacpp、AdaSeq 等
+#- vllm
+  - 心语心言
+  - 医疗
+  - 医疗大模型
+language:
+  - zh
+frameworks: PyTorch
+tasks:
+  - text-generation
+base_model:
+  - Qwen/Qwen3-Next-80B-A3B-Instruct
+base_model_relation: finetune
+metrics:
+  - accuracy
+---
+# DeepPulse-80B TCM Large Model Series
+
+**DeepPulse (深度把脉)** is the core achievement of 心语心言's open-source Traditional Chinese Medicine (TCM) large model series. This series of models uses Qwen3-Next-80B as the base model and has undergone deep fine-tuning using a self-built high-quality TCM clinical medical dataset. This release includes two versions:
+
+* **DeepPulse-80B-Thinking-V0.1**: Focuses on complex clinical reasoning and assisted diagnosis, achieving first place in total score in public evaluations, demonstrating top-tier logical reasoning capabilities in the TCM domain.
+* **DeepPulse-80B-Instruct-V0.1**: Possesses excellent TCM instruction-following capabilities, suitable for a wide range of TCM Q&A and interactive scenarios, with a comprehensive ranking of sixth.
+
+# Public TCM Benchmark Metrics Comparison (MedBench - TCM-5CEval)
+
+TCM-5CEval is an authoritative evaluation benchmark for TCM large models, comprising the following five subtasks that comprehensively assess the model's TCM capabilities:
+
+* **TCM-Exam (中医考试)**: Evaluates the mastery and application of fundamental TCM theories (Yin-Yang, Zang-Fu organs, etc.) and diagnostics knowledge.
+* **TCM-LitQA (典籍问答)**: Tests deep understanding and reasoning of classic TCM texts such as "Huangdi Neijing" and "Shanghan Lun".
+* **TCM-MRCD (临床诊疗)**: Simulates real clinical scenarios, evaluating the model's ability to analyze medical cases, perform pattern differentiation, and make prescription decisions.
+* **TCM-CMM (中药方剂)**: Measures the model's knowledge of Chinese materia medica properties, effects, compatibility contraindications, and formula applications.
+* **TCM-ClinNPT (非药物疗法)**: Assesses ability in acupoint selection for acupuncture, Tuina massage techniques, and pattern-based treatment for specific clinical scenarios.
+
+| No. | Model Name | Organization/Team Name | Release Date | Type | Parameters | Total Score | TCM-Exam | TCM-LitQA | TCM-MRCD | TCM-CMM | TCM-ClinNPT |
+| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
+| 1 | <font color="red">DeepPulse-80B-Thinking-V0.1</font> | <font color="red">心语心言</font> | <font color="red">2025/12/23</font> | <font color="red">开源</font> | <font color="red">80B</font> | <font color="red">71.3</font> | <font color="red">83.0</font> | <font color="red">45.5</font> | <font color="red">75.4</font> | <font color="red">84.9</font> | <font color="red">67.6</font> |
+| 2 | HKR_TCM_HW_v1 | 港仔机器人主动健管团队 | 2025/12/12 | 闭源 | 671B | 70.8 | 85.4 | 44.2 | 73.1 | 83.8 | 67.5 |
+| 3 | Gemini-2.5-Pro-nothinking | Google | 2025/03/25 | 闭源 | N/A | 69.2 | 77.9 | 62.0 | 72.4 | 72.6 | 61.2 |
+| 4 | DeepSeek-V3.2 | DeepSeek | 2025/12/01 | 开源 | 671B | 66.8 | 74.5 | 44.4 | 66.8 | 80.0 | 68.3 |
+| 5 | Grok-4 | xAI | 2025/07/09 | 闭源 | N/A | 66.6 | 73.0 | 59.3 | 68.4 | 68.0 | 64.2 |
+| 6 | <font color="red">DeepPulse-80B-Instruct-V0.1</font> | <font color="red">心语心言</font> | <font color="red">2025/12/23</font> | <font color="red">开源</font> | <font color="red">80B</font> | <font color="red">66.2</font> | <font color="red">74.4</font> | <font color="red">40.7</font> | <font color="red">70.6</font> | <font color="red">79.7</font> | <font color="red">65.6</font> |
+| 7 | Qwen3-235B-A22B-Thinking-2507 | Alibaba | 2025/08/17 | 开源 | 235B | 64.8 | 75.5 | 40.3 | 68.5 | 78.2 | 61.5 |
+| 8 | Claude-Sonnet-4.5 | Anthropic | 2025/09/29 | 闭源 | N/A | 64.8 | 69.8 | 59.3 | 67.2 | 71.7 | 56.0 |
+| 9 | GPT-5 | OpenAI | 2025/08/07 | 闭源 | N/A | 63.6 | 75.0 | 51.9 | 64.1 | 66.6 | 60.6 |
+| 10 | Qwen3-Next-80B-A3B-Thinking | Alibaba | 2025/09/15 | 开源 | 80B | 63.5 | 76.0 | 38.2 | 66.2 | 77.9 | 59.4 |
+| 11 | Llama-4-maverick | Meta | 2025/04/06 | 开源 | 400B | 57.2 | 72.1 | 51.3 | 63.8 | 54.4 | 44.3 |
+| 12 | GPT-4o | OpenAI | 2025/05/13 | 闭源 | 200B | 55.9 | 66.5 | 46.9 | 60.9 | 57.1 | 47.9 |
+
+> Note: "N/A" in the Parameters column indicates that the model's parameter count has not been publicly disclosed.
+>
+> Except for `DeepSeek-V3.2`, `Qwen3-235B-A22B-Thinking-2507`, `Qwen3-Next-80B-A3B-Thinking` which are self-tested deployment data, other models reference publicly available leaderboard data.
+>
+> TCM-5CEval: https://medbench.opencompass.org.cn/track-detail/tcmeval

README.md

@@ -0,0 +1,79 @@
+---
+license: apache-2.0
+tags:
+  - 中医大模型
+
+#model-type:
+##如 gpt、phi、llama、chatglm、baichuan 等
+#- gpt
+
+#domain:
+##如 nlp、cv、audio、multi-modal
+#- nlp
+
+#language:
+##语言代码列表 https://help.aliyun.com/document_detail/215387.html?spm=a2c4g.11186623.0.0.9f8d7467kni6Aa
+#- cn
+
+#metrics:
+##如 CIDEr、Blue、ROUGE 等
+#- CIDEr
+
+#tags:
+##各种自定义，包括 pretrained、fine-tuned、instruction-tuned、RL-tuned 等训练方法和其他
+#- pretrained
+
+#tools:
+##如 vllm、fastchat、llamacpp、AdaSeq 等
+#- vllm
+  - 心语心言
+  - 医疗
+  - 医疗大模型
+language:
+  - zh
+frameworks: PyTorch
+tasks:
+  - text-generation
+base_model:
+  - Qwen/Qwen3-Next-80B-A3B-Instruct
+base_model_relation: finetune
+metrics:
+  - accuracy
+---
+# DeepPulse-80B TCM Large Model Series
+
+**DeepPulse (深度把脉)** is the core achievement of 心语心言's open-source Traditional Chinese Medicine (TCM) large model series. This series of models uses Qwen3-Next-80B as the base model and has undergone deep fine-tuning using a self-built high-quality TCM clinical medical dataset. This release includes two versions:
+
+* **DeepPulse-80B-Thinking-V0.1**: Focuses on complex clinical reasoning and assisted diagnosis, achieving first place in total score in public evaluations, demonstrating top-tier logical reasoning capabilities in the TCM domain.
+* **DeepPulse-80B-Instruct-V0.1**: Possesses excellent TCM instruction-following capabilities, suitable for a wide range of TCM Q&A and interactive scenarios, with a comprehensive ranking of sixth.
+
+# Public TCM Benchmark Metrics Comparison (MedBench - TCM-5CEval)
+
+TCM-5CEval is an authoritative evaluation benchmark for TCM large models, comprising the following five subtasks that comprehensively assess the model's TCM capabilities:
+
+* **TCM-Exam (中医考试)**: Evaluates the mastery and application of fundamental TCM theories (Yin-Yang, Zang-Fu organs, etc.) and diagnostics knowledge.
+* **TCM-LitQA (典籍问答)**: Tests deep understanding and reasoning of classic TCM texts such as "Huangdi Neijing" and "Shanghan Lun".
+* **TCM-MRCD (临床诊疗)**: Simulates real clinical scenarios, evaluating the model's ability to analyze medical cases, perform pattern differentiation, and make prescription decisions.
+* **TCM-CMM (中药方剂)**: Measures the model's knowledge of Chinese materia medica properties, effects, compatibility contraindications, and formula applications.
+* **TCM-ClinNPT (非药物疗法)**: Assesses ability in acupoint selection for acupuncture, Tuina massage techniques, and pattern-based treatment for specific clinical scenarios.
+
+| No. | Model Name | Organization/Team Name | Release Date | Type | Parameters | Total Score | TCM-Exam | TCM-LitQA | TCM-MRCD | TCM-CMM | TCM-ClinNPT |
+| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
+| 1 | <font color="red">DeepPulse-80B-Thinking-V0.1</font> | <font color="red">心语心言</font> | <font color="red">2025/12/23</font> | <font color="red">开源</font> | <font color="red">80B</font> | <font color="red">71.3</font> | <font color="red">83.0</font> | <font color="red">45.5</font> | <font color="red">75.4</font> | <font color="red">84.9</font> | <font color="red">67.6</font> |
+| 2 | HKR_TCM_HW_v1 | 港仔机器人主动健管团队 | 2025/12/12 | 闭源 | 671B | 70.8 | 85.4 | 44.2 | 73.1 | 83.8 | 67.5 |
+| 3 | Gemini-2.5-Pro-nothinking | Google | 2025/03/25 | 闭源 | N/A | 69.2 | 77.9 | 62.0 | 72.4 | 72.6 | 61.2 |
+| 4 | DeepSeek-V3.2 | DeepSeek | 2025/12/01 | 开源 | 671B | 66.8 | 74.5 | 44.4 | 66.8 | 80.0 | 68.3 |
+| 5 | Grok-4 | xAI | 2025/07/09 | 闭源 | N/A | 66.6 | 73.0 | 59.3 | 68.4 | 68.0 | 64.2 |
+| 6 | <font color="red">DeepPulse-80B-Instruct-V0.1</font> | <font color="red">心语心言</font> | <font color="red">2025/12/23</font> | <font color="red">开源</font> | <font color="red">80B</font> | <font color="red">66.2</font> | <font color="red">74.4</font> | <font color="red">40.7</font> | <font color="red">70.6</font> | <font color="red">79.7</font> | <font color="red">65.6</font> |
+| 7 | Qwen3-235B-A22B-Thinking-2507 | Alibaba | 2025/08/17 | 开源 | 235B | 64.8 | 75.5 | 40.3 | 68.5 | 78.2 | 61.5 |
+| 8 | Claude-Sonnet-4.5 | Anthropic | 2025/09/29 | 闭源 | N/A | 64.8 | 69.8 | 59.3 | 67.2 | 71.7 | 56.0 |
+| 9 | GPT-5 | OpenAI | 2025/08/07 | 闭源 | N/A | 63.6 | 75.0 | 51.9 | 64.1 | 66.6 | 60.6 |
+| 10 | Qwen3-Next-80B-A3B-Thinking | Alibaba | 2025/09/15 | 开源 | 80B | 63.5 | 76.0 | 38.2 | 66.2 | 77.9 | 59.4 |
+| 11 | Llama-4-maverick | Meta | 2025/04/06 | 开源 | 400B | 57.2 | 72.1 | 51.3 | 63.8 | 54.4 | 44.3 |
+| 12 | GPT-4o | OpenAI | 2025/05/13 | 闭源 | 200B | 55.9 | 66.5 | 46.9 | 60.9 | 57.1 | 47.9 |
+
+> Note: "N/A" in the Parameters column indicates that the model's parameter count has not been publicly disclosed.
+>
+> Except for `DeepSeek-V3.2`, `Qwen3-235B-A22B-Thinking-2507`, `Qwen3-Next-80B-A3B-Thinking` which are self-tested deployment data, other models reference publicly available leaderboard data.
+>
+> TCM-5CEval: https://medbench.opencompass.org.cn/track-detail/tcmeval

added_tokens.json

@@ -0,0 +1,28 @@
+{
+  "</think>": 151668,
+  "</tool_call>": 151658,
+  "</tool_response>": 151666,
+  "<think>": 151667,
+  "<tool_call>": 151657,
+  "<tool_response>": 151665,
+  "<|box_end|>": 151649,
+  "<|box_start|>": 151648,
+  "<|endoftext|>": 151643,
+  "<|file_sep|>": 151664,
+  "<|fim_middle|>": 151660,
+  "<|fim_pad|>": 151662,
+  "<|fim_prefix|>": 151659,
+  "<|fim_suffix|>": 151661,
+  "<|im_end|>": 151645,
+  "<|im_start|>": 151644,
+  "<|image_pad|>": 151655,
+  "<|object_ref_end|>": 151647,
+  "<|object_ref_start|>": 151646,
+  "<|quad_end|>": 151651,
+  "<|quad_start|>": 151650,
+  "<|repo_name|>": 151663,
+  "<|video_pad|>": 151656,
+  "<|vision_end|>": 151653,
+  "<|vision_pad|>": 151654,
+  "<|vision_start|>": 151652
+}

chat_template.jinja

@@ -0,0 +1,61 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- for message in messages %}
+    {%- if message.content is string %}
+        {%- set content = message.content %}
+    {%- else %}
+        {%- set content = '' %}
+    {%- endif %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+{%- endif %}

config.json

@@ -0,0 +1,94 @@
+{
+  "architectures": [
+    "Qwen3NextForCausalLM"
+  ],
+  "attention_bias": false,
+  "attention_dropout": 0.0,
+  "bos_token_id": 151643,
+  "decoder_sparse_step": 1,
+  "dtype": "bfloat16",
+  "eos_token_id": 151645,
+  "full_attention_interval": 4,
+  "head_dim": 256,
+  "hidden_act": "silu",
+  "hidden_size": 2048,
+  "initializer_range": 0.02,
+  "intermediate_size": 5120,
+  "layer_types": [
+    "linear_attention",
+    "linear_attention",
+    "linear_attention",
+    "full_attention",
+    "linear_attention",
+    "linear_attention",
+    "linear_attention",
+    "full_attention",
+    "linear_attention",
+    "linear_attention",
+    "linear_attention",
+    "full_attention",
+    "linear_attention",
+    "linear_attention",
+    "linear_attention",
+    "full_attention",
+    "linear_attention",
+    "linear_attention",
+    "linear_attention",
+    "full_attention",
+    "linear_attention",
+    "linear_attention",
+    "linear_attention",
+    "full_attention",
+    "linear_attention",
+    "linear_attention",
+    "linear_attention",
+    "full_attention",
+    "linear_attention",
+    "linear_attention",
+    "linear_attention",
+    "full_attention",
+    "linear_attention",
+    "linear_attention",
+    "linear_attention",
+    "full_attention",
+    "linear_attention",
+    "linear_attention",
+    "linear_attention",
+    "full_attention",
+    "linear_attention",
+    "linear_attention",
+    "linear_attention",
+    "full_attention",
+    "linear_attention",
+    "linear_attention",
+    "linear_attention",
+    "full_attention"
+  ],
+  "linear_conv_kernel_dim": 4,
+  "linear_key_head_dim": 128,
+  "linear_num_key_heads": 16,
+  "linear_num_value_heads": 32,
+  "linear_value_head_dim": 128,
+  "max_position_embeddings": 262144,
+  "mlp_only_layers": [],
+  "model_type": "qwen3_next",
+  "moe_intermediate_size": 512,
+  "norm_topk_prob": true,
+  "num_attention_heads": 16,
+  "num_experts": 512,
+  "num_experts_per_tok": 10,
+  "num_hidden_layers": 48,
+  "num_key_value_heads": 2,
+  "output_router_logits": false,
+  "partial_rotary_factor": 0.25,
+  "rms_norm_eps": 1e-06,
+  "rope_scaling": null,
+  "rope_theta": 10000000,
+  "router_aux_loss_coef": 0.001,
+  "shared_expert_intermediate_size": 512,
+  "tie_word_embeddings": false,
+  "transformers_version": "4.57.1",
+  "use_cache": true,
+  "use_sliding_window": false,
+  "vocab_size": 151936
+}

model17_0.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0b224fe5b651e0b0da8b1208776bb86d18690f4ff6a2d1d85a671a89beaca3f6
+size 4832125960

model18_0.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8fe7a9d44b35f45564a134b8888a3170ad830cddc08e9c1e7fced650f5e1ee76
+size 4832125960

model19_0.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2d4facec77c41938e366e7ff290cfe9934362d3dfb5688b5ec167351988b45a7
+size 4832125960

model1_0.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:13dedd057368d95a8a03539b3bbd14ad013aa91309e468577cbc572227a9a864
+size 4832124808

model7_0.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9866a670e13ef5324617f069ed411b666488de7cd6b6e8287b6c490ed9a0cfff
+size 4832124808

special_tokens_map.json

@@ -0,0 +1,31 @@
+{
+  "additional_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "eos_token": {
+    "content": "<|im_end|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  }
+}