[usability] add qwen2.5 (qwq) template

docs/source/examples/supported_conversation_template.md

@@ -13,6 +13,7 @@
   - [Mixtral 8x7B](#mixtral-8x7b)
   - [Phi-3](#phi-3)
   - [Qwen-2](#qwen-2)
+  - [Qwen-2.5](#qwen-25)
   - [Yi](#yi)
   - [Yi-1.5](#yi-15)
   - [Zephyr](#zephyr)
@@ -405,6 +406,95 @@ The conversation template for Mixtral 8x7B is slightly different from the templa
 ```
 
 
+## Qwen-2.5
+
+(Also QwQ)
+
+**With a system message**
+```
+<|im_start|>system\n{{system_message}}<|im_end|>\n<|im_start|>user\n{{user_message_0}}<|im_end|>\n
+```
+
+**Without a system message**
+```
+<|im_start|>user\n{{user_message_0}}<|im_end|>\n
+```
+
+**A complete conversation**
+```
+<|im_start|>system\n{{system_message}}<|im_end|>\n<|im_start|>user\n{{user_message_0}}<|im_end|>\n<|im_start|>assistant\n{{assistant_reply_0}}<|im_end|>\n
+```
+
+**Multiple rounds**
+```
+<|im_start|>system\n{{system_message}}<|im_end|>\n<|im_start|>user\n{{user_message_0}}<|im_end|>\n<|im_start|>assistant\n{{assistant_reply_0}}<|im_end|>\n<|im_start|>user\n{{user_message_1}}<|im_end|>\n<|im_start|>assistant\n{{assistant_reply_1}}<|im_end|>\n
+```
+
+**jinja template**  
+[[Reference]()]
+```
+{%- if tools %}
+    {{- '<|im_start|>system\\n' }}
+    {%- if messages[0]['role'] == 'system' %}
+        {{- messages[0]['content'] }}
+    {%- else %}
+        {{- 'You are a helpful and harmless assistant. You are Qwen developed by Alibaba. You should think step-by-step.' }}
+    {%- endif %}
+    {{- \"\\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>\" }}
+    {%- 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' }}
+    {%- else %}
+        {{- '<|im_start|>system\\nYou are a helpful and harmless assistant. You are Qwen developed by Alibaba. You should think step-by-step.<|im_end|>\\n' }}
+    {%- endif %}
+{%- endif %}
+{%- for message in messages %}
+    {%- if (message.role == \"user\") or (message.role == \"system\" and not loop.first) or (message.role == \"assistant\" and not message.tool_calls) %}
+        {{- '<|im_start|>' + message.role + '\\n' + message.content + '<|im_end|>' + '\\n' }}
+    {%- elif message.role == \"assistant\" %}
+        {{- '<|im_start|>' + message.role }}
+        {%- if message.content %}
+            {{- '\\n' + message.content }}
+        {%- endif %}
+        {%- for tool_call in message.tool_calls %}
+            {%- if tool_call.function is defined %}
+                {%- set tool_call = tool_call.function %}
+            {%- endif %}
+            {{- '\\n<tool_call>\\n{\"name\": \"' }}
+            {{- tool_call.name }}
+            {{- '\", \"arguments\": ' }}
+            {{- tool_call.arguments | tojson }}
+            {{- '}\\n</tool_call>' }}
+        {%- endfor %}
+        {{- '<|im_end|>\\n' }}
+    {%- elif message.role == \"tool\" %}
+        {%- if (loop.index0 == 0) or (messages[loop.index0 - 1].role != \"tool\") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\\n<tool_response>\\n' }}
+        {{- message.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 %}
+```
+
+**Filled Example**
+```
+<|im_start|>system\nYou are a chatbot developed by LMFlow team.<|im_end|>\n<|im_start|>user\nWho are you?<|im_end|>\n<|im_start|>assistant\nI am a chatbot developed by LMFlow team.<|im_end|>\n<|im_start|>user\nHow old are you?<|im_end|>\n<|im_start|>assistant\nI don't age like humans do. I exist as a piece of software, so I don't have a concept of age in the traditional sense.<|im_end|>\n
+```
+
+
 ## Yi
 **With a system message**
 ```

src/lmflow/utils/conversation_template/__init__.py

@@ -10,7 +10,7 @@
 from .internlm import INTERNLM2_TEMPLATE
 from .llama import LLAMA2_TEMPLATE, LLAMA3_TEMPLATE, LLAMA3_TEMPLATE_FOR_TOOL
 from .phi import PHI3_TEMPLATE
-from .qwen import QWEN2_TEMPLATE, QWEN2_TEMPLATE_FOR_TOOL
+from .qwen import QWEN2_TEMPLATE, QWEN2_TEMPLATE_FOR_TOOL, QWEN2_5_TEMPLATE
 from .yi import YI1_5_TEMPLATE
 from .zephyr import ZEPHYR_TEMPLATE
 
@@ -31,6 +31,7 @@
     'phi3': PHI3_TEMPLATE,
     'qwen2': QWEN2_TEMPLATE,
     'qwen2_for_tool': QWEN2_TEMPLATE_FOR_TOOL,
+    'qwen2_5': QWEN2_5_TEMPLATE,
     'yi': CHATML_TEMPLATE,
     'yi1_5': YI1_5_TEMPLATE,
     'zephyr': ZEPHYR_TEMPLATE

src/lmflow/utils/conversation_template/qwen.py

@@ -1,6 +1,10 @@
 #!/usr/bin/env python
 # coding=utf-8
 # Copyright 2024 Statistics and Machine Learning Research Group. All rights reserved.
+from typing import Dict, Set, Sequence, Literal, Union, List, Optional, Tuple
+
+from transformers import PreTrainedTokenizer
+
 from .base import StringFormatter, TemplateComponent, ConversationTemplate, ConversationTemplateForTool
 
 
@@ -24,6 +28,7 @@
     separator=TemplateComponent(type='string', content='\n')
 )
 
+
 QWEN2_TEMPLATE_FOR_TOOL = ConversationTemplateForTool(
     template_name='qwen2_for_tool',
     user_formatter=StringFormatter(
@@ -52,4 +57,51 @@
         ]
     ),
     separator=TemplateComponent(type='string', content='\n')
+)
+
+
+class Qwen2_5ConversationTemplate(ConversationTemplateForTool):
+    def _handle_tools(self, tools: Optional[List[str]]) -> str:
+        prompt1 = "\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>"
+        prompt2 = '\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>'
+        tools_out = ''
+
+        if tools is not None:
+            tools_out = prompt1
+            for tool in tools:
+                tools_out += "\n" + tool
+            tools_out += prompt2
+
+        return tools_out
+
+
+QWEN2_5_TEMPLATE = Qwen2_5ConversationTemplate(
+    template_name='qwen2_5',
+    user_formatter=StringFormatter(
+        template=[
+            TemplateComponent(type='string', content='<|im_start|>user\n{{content}}<|im_end|>\n')
+        ]
+    ),
+    function_formatter=StringFormatter(
+        template=[
+            TemplateComponent(type='string', content='<|im_start|>assistant\n{{content}}<|im_end|>\n')
+        ]
+    ),
+    observation_formatter=StringFormatter(
+        template=[
+            TemplateComponent(type='string', content='<|im_start|>tool\n{{content}}<|im_end|>\n')
+        ]
+    ),
+    assistant_formatter=StringFormatter(
+        template=[
+            TemplateComponent(type='string', content='<|im_start|>assistant\n{{content}}<|im_end|>\n')
+        ]
+    ),
+    system_formatter=StringFormatter(
+        template=[
+            TemplateComponent(type='string', content='<|im_start|>system\n{{content}}<|im_end|>\n')
+        ]
+    ),
+    separator=TemplateComponent(type='string', content='\n'),
+    system_default='You are a helpful and harmless assistant. You are Qwen developed by Alibaba. You should think step-by-step.'
 )