更新lmdeoploy版本

Author: shell-nlp

gpt_server/model_backend/lmdeploy_backend.py

@@ -9,7 +9,7 @@
 from typing import Any, Dict, AsyncGenerator, List, Optional
 from lmdeploy.archs import get_task
 from gpt_server.model_handler.reasoning_parser import ReasoningParserManager
-from lmdeploy.serve.async_engine import get_names_from_model
+from lmdeploy.serve.async_engine import best_match_model
 from loguru import logger
 from gpt_server.model_backend.base import ModelBackend
 from gpt_server.settings import get_model_config
@@ -124,7 +124,7 @@ def __init__(self, model_path, tokenizer: PreTrainedTokenizerBase) -> None:
             backend=backend,
             backend_config=backend_config,
         )
-        model_name, chat_template_name = get_names_from_model(model_path=model_path)
+        chat_template_name = best_match_model(query=model_path)
         self.chat_template_name = chat_template_name
         self.tokenizer = self.async_engine.tokenizer
         self.reasoning_parser_cache = {}
@@ -166,18 +166,14 @@ async def stream_chat(self, params: Dict[str, Any]) -> AsyncGenerator:
             skip_special_tokens=True,
             response_format=params["response_format"],
         )
-        if params.get("tools", None) or is_messages_with_tool(messages=messages):
-            messages = prompt or messages  # 解决lmdeploy 的提示模板不支持 tools
+        # if params.get("tools", None) or is_messages_with_tool(messages=messages):
+        #     messages = prompt or messages  # 解决lmdeploy 的提示模板不支持 tools
         logger.info(f"chat_template_name: {self.chat_template_name}")
-        if self.chat_template_name == "base":
-            messages = prompt or messages
-        multimodal = params.get("multimodal", False)
-        if multimodal:  # 多模态模型
-            messages = params["messages"]
-        if isinstance(messages, str):
-            logger.info(f"使用prompt模式")
-        else:
-            logger.info(f"使用messages模式")
+        # if self.chat_template_name == "base":
+        #     messages = prompt or messages
+        # multimodal = params.get("multimodal", False)
+        # if multimodal:  # 多模态模型
+        #     messages = params["messages"]
         results_generator = self.async_engine.generate(
             messages=messages,
             session_id=int(request_id),

gpt_server/model_backend/vllm_backend.py

@@ -61,7 +61,6 @@ def __init__(self, model_path, tokenizer: AutoTokenizer) -> None:
     async def stream_chat(self, params: Dict[str, Any]) -> AsyncGenerator:
         prompt = params.get("prompt", "")
         messages = params["messages"]
-        logger.info(f"prompt：\n{prompt}")
         request_id = params.get("request_id", "0")
         temperature = float(params.get("temperature", 0.8))
         top_p = float(params.get("top_p", 0.8))
@@ -74,6 +73,7 @@ async def stream_chat(self, params: Dict[str, Any]) -> AsyncGenerator:
         repetition_penalty = float(params.get("repetition_penalty", 1.0))
         enable_thinking = bool(params.get("enable_thinking", True))
         request = params.get("request", None)
+        tools = params.get("tools", None)
         # Handle stop_str
         stop = set()
         if isinstance(stop_str, str) and stop_str != "":
@@ -89,7 +89,7 @@ async def stream_chat(self, params: Dict[str, Any]) -> AsyncGenerator:
             conversation, mm_data_future = parse_chat_messages_futures(
                 messages, model_config, tokenizer, content_format="string"
             )
-            tools = params.get("tools", None)
+
             prompt = apply_hf_chat_template(
                 tokenizer,
                 conversation=conversation,
@@ -102,11 +102,22 @@ async def stream_chat(self, params: Dict[str, Any]) -> AsyncGenerator:
             mm_data = await mm_data_future
             inputs = {"multi_modal_data": mm_data, "prompt": prompt}
         else:
+            conversation = messages
+            prompt = apply_hf_chat_template(
+                tokenizer,
+                conversation=conversation,
+                chat_template=tokenizer.get_chat_template(),
+                add_generation_prompt=True,
+                tools=tools,
+                model_config=await self.engine.get_model_config(),
+                enable_thinking=enable_thinking,
+            )
             input_ids = params.get("input_ids", None)
             inputs = {"prompt": prompt}
             if input_ids is not None:
                 prompt_token_ids = input_ids.tolist()[0]
                 inputs["prompt_token_ids"] = prompt_token_ids
+        logger.info(f"prompt：\n{prompt}")
         # ----------------------------------------------------------------
         # make sampling params in vllm
         top_p = max(top_p, 1e-5)

gpt_server/model_handler/chat_template/get_chat_template.py

@@ -0,0 +1,33 @@
+from pathlib import Path
+from typing import Literal
+
+cur_path = Path(__file__).parent
+
+
+def get_chat_template(model_name: str = "", lang: Literal["en", "zh"] = "en") -> str:
+    """获取chat_template
+
+    Parameters
+    ----------
+    model_name : str
+        模型名称
+    lang : str, optional
+        语言, by default en
+
+    Returns
+    -------
+    str
+        chat_template
+    """
+    suffix = ""
+    if lang == "zh":
+        suffix = "_zh"
+    if model_name in ["qwen3", "qwen2_5", "qwen"]:
+        with open(cur_path / f"qwen3{suffix}.jinja", "r", encoding="utf8") as f:
+            return f.read()
+
+
+if __name__ == "__main__":
+
+    chat_template = get_chat_template("qwen3", lang="zh")
+    print(chat_template)

gpt_server/model_handler/chat_template/qwen3.jinja

@@ -0,0 +1,88 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- else %}
+        {{- 'You are a helpful assistant. \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 %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and message.content is string and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- 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" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- 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 %}

gpt_server/model_handler/chat_template/qwen3_zh.jinja

@@ -0,0 +1,88 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- else %}
+        {{- 'You are a helpful assistant. \n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\n您可以调用一个或多个function来协助处理用户查询。\n\n在<tools></tools> XML标签中提供了function的签名(即函数的结构信息):\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\n对于单个function的调用, 返回一个包含function name和参数的 JSON 对象，并用 <tool_call></tool_call> XML 标签包裹,形如:\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 %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and message.content is string and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- 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" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- 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 %}

gpt_server/model_handler/prompts.py

@@ -1,10 +1,15 @@
 from typing import Optional
-from lmdeploy.model import MODELS, Qwen7BChat, ChatGLM3, get_text
+from lmdeploy.model import (
+    MODELS,
+    get_text,
+    HFChatTemplate,
+    BaseChatTemplate,
+)
 import json
 
 
 @MODELS.register_module(name="glm4", force=True)
-class Glm4Chat(ChatGLM3):
+class Glm4Chat(BaseChatTemplate):
     """Chat template of glm-4 model."""
 
     def __init__(
@@ -125,7 +130,7 @@ def messages2prompt_base(self, messages, sequence_start=True, tools=None, **kwar
 
 
 @MODELS.register_module(name="qwen2_5")
-class Qwen2d5Chat(Qwen7BChat):
+class Qwen2d5Chat(BaseChatTemplate):
     """Chat template for Qwen2.5-Instruct series."""
 
     def __init__(
@@ -253,15 +258,15 @@ def match(cls, model_path: str) -> Optional[str]:
 if __name__ == "__main__":
     chat_template = MODELS.module_dict["qwen2_5"]()
     messages = [
-        {"role": "system", "content": "我的Qwen "},
+        # {"role": "system", "content": "我的Qwen "},
         {"role": "user", "content": "你是谁 "},
     ]
     tools = [
         {
             "type": "function",
             "function": {
                 "name": "get_weather",
-                "description": "Get the current weather in a given location",
+                "description": "在给定位置获得当前的天气",
                 "parameters": {
                     "type": "object",
                     "properties": {
@@ -296,4 +301,28 @@ def match(cls, model_path: str) -> Optional[str]:
     ]
     # tools = None
     prompt = chat_template.messages2prompt(messages, True, tools)
+    # print(prompt)
+    #
+    with open(
+        "/home/dev/liuyu/project/gpt_server/gpt_server/model_handler/chat_template/qwen3_zh.jinja",
+        "r",
+        encoding="utf8",
+    ) as f:
+        qwen3_next = f.read()
+
+    # qwen2_5xxx = "{%- if tools %}\n    {{- '<|im_start|>system\\n' }}\n    {%- if messages[0]['role'] == 'system' %}\n        {{- messages[0]['content'] }}\n    {%- else %}\n        {{- 'You are Qwen, created by Alibaba Cloud. You are a helpful assistant.' }}\n    {%- endif %}\n    {{- \"\\n\\n# 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>\" }}\n    {%- for tool in tools %}\n        {{- \"\\n\" }}\n        {{- tool | tojson }}\n    {%- endfor %}\n    {{- \"\\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\" }}\n{%- else %}\n    {%- if messages[0]['role'] == 'system' %}\n        {{- '<|im_start|>system\\n' + messages[0]['content'] + '<|im_end|>\\n' }}\n    {%- else %}\n        {{- '<|im_start|>system\\nYou are Qwen, created by Alibaba Cloud. You are a helpful assistant.<|im_end|>\\n' }}\n    {%- endif %}\n{%- endif %}\n{%- for message in messages %}\n    {%- if (message.role == \"user\") or (message.role == \"system\" and not loop.first) or (message.role == \"assistant\" and not message.tool_calls) %}\n        {{- '<|im_start|>' + message.role + '\\n' + message.content + '<|im_end|>' + '\\n' }}\n    {%- elif message.role == \"assistant\" %}\n        {{- '<|im_start|>' + message.role }}\n        {%- if message.content %}\n            {{- '\\n' + message.content }}\n        {%- endif %}\n        {%- for tool_call in message.tool_calls %}\n            {%- if tool_call.function is defined %}\n                {%- set tool_call = tool_call.function %}\n            {%- endif %}\n            {{- '\\n<tool_call>\\n{\"name\": \"' }}\n            {{- tool_call.name }}\n            {{- '\", \"arguments\": ' }}\n            {{- tool_call.arguments | tojson }}\n            {{- '}\\n</tool_call>' }}\n        {%- endfor %}\n        {{- '<|im_end|>\\n' }}\n    {%- elif message.role == \"tool\" %}\n        {%- if (loop.index0 == 0) or (messages[loop.index0 - 1].role != \"tool\") %}\n            {{- '<|im_start|>user' }}\n        {%- endif %}\n        {{- '\\n<tool_response>\\n' }}\n        {{- message.content }}\n        {{- '\\n</tool_response>' }}\n        {%- if loop.last or (messages[loop.index0 + 1].role != \"tool\") %}\n            {{- '<|im_end|>\\n' }}\n        {%- endif %}\n    {%- endif %}\n{%- endfor %}\n{%- if add_generation_prompt %}\n    {{- '<|im_start|>assistant\\n' }}\n{%- endif %}\n"
+    # print(qwen3_next)
+    # assert 0
+    from transformers import AutoTokenizer
+
+    model_path = "/home/dev/model/Qwen/Qwen2___5-32B-Instruct-AWQ/"
+    tokenizer = AutoTokenizer.from_pretrained(model_path, trust_remote_code=True)
+    # chat_template
+    prompt = tokenizer.apply_chat_template(
+        messages,
+        tokenize=False,
+        add_generation_prompt=True,
+        chat_template=qwen3_next,
+        tools=tools,
+    )
     print(prompt)