xinference 0.14.4.post1__py3-none-any.whl → 0.15.1__py3-none-any.whl

xinference/model/llm/llm_family.py

@@ -52,7 +52,7 @@ from . import LLM
 logger = logging.getLogger(__name__)
 
 DEFAULT_CONTEXT_LENGTH = 2048
-BUILTIN_LLM_PROMPT_STYLE: Dict[str, "PromptStyleV1"] = {}
+BUILTIN_LLM_PROMPT_STYLE: Dict[str, Dict[str, Any]] = {}
 BUILTIN_LLM_MODEL_CHAT_FAMILIES: Set[str] = set()
 BUILTIN_LLM_MODEL_GENERATE_FAMILIES: Set[str] = set()
 BUILTIN_LLM_MODEL_TOOL_CALL_FAMILIES: Set[str] = set()
@@ -127,32 +127,24 @@ class MLXLLMSpecV1(BaseModel):
         return v
 
 
-class PromptStyleV1(BaseModel):
-    style_name: str
-    system_prompt: str = ""
-    roles: List[str]
-    intra_message_sep: str = ""
-    inter_message_sep: str = ""
-    stop: Optional[List[str]]
-    stop_token_ids: Optional[List[int]]
-
-
 class LLMFamilyV1(BaseModel):
     version: Literal[1]
     context_length: Optional[int] = DEFAULT_CONTEXT_LENGTH
     model_name: str
     model_lang: List[str]
-    model_ability: List[Literal["embed", "generate", "chat", "tools", "vision"]]
+    model_ability: List[
+        Literal["embed", "generate", "chat", "tools", "vision", "audio"]
+    ]
     model_description: Optional[str]
     # reason for not required str here: legacy registration
     model_family: Optional[str]
     model_specs: List["LLMSpecV1"]
-    prompt_style: Optional["PromptStyleV1"]
+    chat_template: Optional[str]
+    stop_token_ids: Optional[List[int]]
+    stop: Optional[List[str]]
 
 
 class CustomLLMFamilyV1(LLMFamilyV1):
-    prompt_style: Optional[Union["PromptStyleV1", str]]  # type: ignore
-
     @classmethod
     def parse_raw(
         cls: Any,
@@ -176,6 +168,11 @@ class CustomLLMFamilyV1(LLMFamilyV1):
         except (ValueError, TypeError, UnicodeDecodeError) as e:
             raise ValidationError([ErrorWrapper(e, loc=ROOT_KEY)], cls)
         llm_spec: CustomLLMFamilyV1 = cls.parse_obj(obj)
+        vision_model_names: Set[str] = {
+            family.model_name
+            for family in BUILTIN_LLM_FAMILIES
+            if "vision" in family.model_ability
+        }
 
         # check model_family
         if llm_spec.model_family is None:
@@ -183,61 +180,45 @@ class CustomLLMFamilyV1(LLMFamilyV1):
                 f"You must specify `model_family` when registering custom LLM models."
             )
         assert isinstance(llm_spec.model_family, str)
+        # TODO: Currently, tool call and vision models cannot be registered if it is not the builtin model_family
         if (
-            llm_spec.model_family != "other"
-            and "chat" in llm_spec.model_ability
-            and llm_spec.model_family not in BUILTIN_LLM_MODEL_CHAT_FAMILIES
-        ):
-            raise ValueError(
-                f"`model_family` for chat model must be `other` or one of the following values: \n"
-                f"{', '.join(list(BUILTIN_LLM_MODEL_CHAT_FAMILIES))}"
-            )
-        if (
-            llm_spec.model_family != "other"
-            and "tools" in llm_spec.model_ability
+            "tools" in llm_spec.model_ability
             and llm_spec.model_family not in BUILTIN_LLM_MODEL_TOOL_CALL_FAMILIES
         ):
             raise ValueError(
-                f"`model_family` for tool call model must be `other` or one of the following values: \n"
+                f"`model_family` for tool call model must be one of the following values: \n"
                 f"{', '.join(list(BUILTIN_LLM_MODEL_TOOL_CALL_FAMILIES))}"
             )
         if (
-            llm_spec.model_family != "other"
-            and "chat" not in llm_spec.model_ability
-            and llm_spec.model_family not in BUILTIN_LLM_MODEL_GENERATE_FAMILIES
+            "vision" in llm_spec.model_ability
+            and llm_spec.model_family not in vision_model_names
         ):
             raise ValueError(
-                f"`model_family` for generate model must be `other` or one of the following values: \n"
-                f"{', '.join(list(BUILTIN_LLM_MODEL_GENERATE_FAMILIES))}"
+                f"`model_family` for multimodal model must be one of the following values: \n"
+                f"{', '.join(list(vision_model_names))}"
             )
-        # set prompt style when it is the builtin model family
+        # set chat_template when it is the builtin model family
+        if llm_spec.chat_template is None and "chat" in llm_spec.model_ability:
+            llm_spec.chat_template = llm_spec.model_family
+
+        # handle chat_template when user choose existing model_family
         if (
-            llm_spec.prompt_style is None
-            and llm_spec.model_family != "other"
-            and "chat" in llm_spec.model_ability
+            llm_spec.chat_template is not None
+            and llm_spec.chat_template in BUILTIN_LLM_PROMPT_STYLE
         ):
-            llm_spec.prompt_style = llm_spec.model_family
-
-        # handle prompt style when user choose existing style
-        if llm_spec.prompt_style is not None and isinstance(llm_spec.prompt_style, str):
-            prompt_style_name = llm_spec.prompt_style
-            if prompt_style_name not in BUILTIN_LLM_PROMPT_STYLE:
-                raise ValueError(
-                    f"Xinference does not support the prompt style name: {prompt_style_name}"
-                )
-            llm_spec.prompt_style = BUILTIN_LLM_PROMPT_STYLE[prompt_style_name]
+            llm_spec.stop_token_ids = BUILTIN_LLM_PROMPT_STYLE[llm_spec.chat_template][
+                "stop_token_ids"
+            ]
+            llm_spec.stop = BUILTIN_LLM_PROMPT_STYLE[llm_spec.chat_template]["stop"]
+            llm_spec.chat_template = BUILTIN_LLM_PROMPT_STYLE[llm_spec.chat_template][
+                "chat_template"
+            ]
 
         # check model ability, registering LLM only provides generate and chat
         # but for vision models, we add back the abilities so that
         # gradio chat interface can be generated properly
         if (
-            llm_spec.model_family != "other"
-            and llm_spec.model_family
-            in {
-                family.model_name
-                for family in BUILTIN_LLM_FAMILIES
-                if "vision" in family.model_ability
-            }
+            llm_spec.model_family in vision_model_names
             and "vision" not in llm_spec.model_ability
         ):
             llm_spec.model_ability.append("vision")

xinference/model/llm/llm_family_csghub.json

@@ -43,25 +43,17 @@
         "model_hub": "csghub"
       }
     ],
-    "prompt_style": {
-      "style_name": "QWEN",
-      "system_prompt": "You are a helpful assistant.",
-      "roles": [
-        "user",
-        "assistant"
-      ],
-      "intra_message_sep": "\n",
-      "stop_token_ids": [
-        151643,
-        151644,
-        151645
-      ],
-      "stop": [
-        "<|endoftext|>",
-        "<|im_start|>",
-        "<|im_end|>"
-      ]
-    }
+    "chat_template": "{%- macro json_to_python_type(json_spec) %}\n    {%- set basic_type_map = {\n    \"string\": \"str\",\n    \"number\": \"float\",\n    \"integer\": \"int\",\n    \"boolean\": \"bool\"\n} %}\n    {%- if basic_type_map[json_spec.type] is defined %}\n        {{- basic_type_map[json_spec.type] }}\n    {%- elif json_spec.type == \"array\" %}\n        {{- \"list[\" +  json_to_python_type(json_spec|items) + \"]\" }}\n    {%- elif json_spec.type == \"object\" %}\n        {%- if json_spec.additionalProperties is defined %}\n            {{- \"dict[str, \" + json_to_python_type(json_spec.additionalProperties) + ']' }}\n        {%- else %}\n            {{- \"dict\" }}\n        {%- endif %}\n    {%- elif json_spec.type is iterable %}\n        {{- \"Union[\" }}\n        {%- for t in json_spec.type %}\n            {{- json_to_python_type({\"type\": t}) }}\n            {%- if not loop.last %}\n                {{- \",\" }}\n            {%- endif %}\n        {%- endfor %}\n        {{- \"]\" }}\n    {%- else %}\n        {{- \"Any\" }}\n    {%- endif %}\n{%- endmacro %}\n\n{%- if tools %}\n    {{- '<|im_start|>system\n' }}\n    {%- if messages[0]['role'] == 'system' %}\n        {{- messages[0]['content'] + '\n\n' }}\n    {%- endif %}\n    {{- '# Tools\n\n' }}\n    {{- \"You are a function calling AI model. You are provided with function signatures within <tools></tools> XML tags. You may call one or more functions to assist with the user query. Don't make assumptions about what values to plug into functions. Here are the available tools: <tools> \" }}\n    {%- for tool in tools %}\n        {%- if tool.function is defined %}\n            {%- set tool = tool.function %}\n        {%- endif %}\n        {{- '{\"type\": \"function\", \"function\": ' }}\n        {{- '{\"name\": ' + tool.name + '\", ' }}\n        {{- '\"description\": \"' + tool.name + '(' }}\n        {%- for param_name, param_fields in tool.parameters.properties|items %}\n            {{- param_name + \": \" + json_to_python_type(param_fields) }}\n            {%- if not loop.last %}\n                {{- \", \" }}\n            {%- endif %}\n        {%- endfor %}\n        {{- \")\" }}\n        {%- if tool.return is defined %}\n            {{- \" -> \" + json_to_python_type(tool.return) }}\n        {%- endif %}\n        {{- \" - \" + tool.description + \"\n\n\" }}\n        {%- for param_name, param_fields in tool.parameters.properties|items %}\n            {%- if loop.first %}\n                {{- \"    Args:\n\" }}\n            {%- endif %}\n            {{- \"        \" + param_name + \"(\" + json_to_python_type(param_fields) + \"): \" + param_fields.description|trim }}\n        {%- endfor %}\n        {%- if tool.return is defined and tool.return.description is defined %}\n            {{- \"\n    Returns:\n        \" + tool.return.description }}\n        {%- endif %}\n        {{- '\"' }}\n        {{- ', \"parameters\": ' }}\n        {%- if tool.parameters.properties | length == 0 %}\n            {{- \"{}\" }}\n        {%- else %}\n            {{- tool.parameters|tojson }}\n        {%- endif %}\n        {{- \"}\" }}\n        {%- if not loop.last %}\n            {{- \"\n\" }}\n        {%- endif %}\n    {%- endfor %}\n    {{- \" </tools>\" }}\n    {{- 'Use the following pydantic model json schema for each tool call you will make: {\"properties\": {\"arguments\": {\"title\": \"Arguments\", \"type\": \"object\"}, \"name\": {\"title\": \"Name\", \"type\": \"string\"}}, \"required\": [\"arguments\", \"name\"], \"title\": \"FunctionCall\", \"type\": \"object\"}\n' }}\n    {{- \"For each function call return a json object with function name and arguments within <tool_call></tool_call> XML tags as follows:\n\" }}\n    {{- \"<tool_call>\n\" }}\n    {{- '{\"name\": <function-name>, \"arguments\": <args-json-object>}\n' }}\n    {{- '</tool_call><|im_end|>\n' }}\n{%- else %}\n    {%- if messages[0]['role'] != 'system' %}\n        {{- '<|im_start|>system\nYou are a helpful assistant.<|im_end|>\n' }}\n    {%- else %}\n        {{- '<|im_start|>system\n' + messages[0]['content'] + '<|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 message.tool_calls is not defined) %}\n        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}\n    {%- elif message.role == \"assistant\" %}\n        {{- '<|im_start|>' + message.role + '\n<tool_call>\n' }}\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            {{- '\"name\": \"' }}\n            {{- tool_call.name }}\n            {%- if tool_call.arguments is defined %}\n                {{- ', ' }}\n                {{- '\"arguments\": ' }}\n                {{- tool_call.arguments|tojson }}\n            {%- endif %}\n            {{- '\"}' }}\n            {{- '\n</tool_call>' }}\n        {%- endfor %}\n        {{- '<|im_end|>\n' }}\n    {%- elif message.role == \"tool\" %}\n        {%- if not message.name is defined %}\n            {{- raise_exception(\"Tool response dicts require a 'name' key indicating the name of the called function!\") }}\n        {%- endif %}\n        {{- '<|im_start|>user\n<tool_response>\n' }}\n        {{- '{\"name\": \"' }}\n        {{- message.name }}\n        {{- '\", \"content\": ' }}\n        {{- message.content|tojson + '}' }}\n        {{- '\n</tool_response><|im_end|>\n' }}\n    {%- endif %}\n{%- endfor %}\n{%- if add_generation_prompt %}\n    {{- '<|im_start|>assistant\n' }}\n{%- endif %}",
+    "stop_token_ids": [
+      151643,
+      151644,
+      151645
+    ],
+    "stop": [
+      "<|endoftext|>",
+      "<|im_start|>",
+      "<|im_end|>"
+    ]
   },
   {
     "version": 1,
@@ -85,21 +77,12 @@
         "model_hub": "csghub"
       }
     ],
-    "prompt_style": {
-      "style_name": "NO_COLON_TWO",
-      "system_prompt": "<|system|>\nYou are a creative super artificial intelligence assistant, possessing all the knowledge of humankind. Your name is csg-wukong, developed by OpenCSG. You need to understand and infer the true intentions of users based on the topics discussed in the chat history, and respond to user questions correctly as required. You enjoy responding to users with accurate and insightful answers. Please pay attention to the appropriate style and format when replying, try to avoid repetitive words and sentences, and keep your responses as concise and profound as possible. You carefully consider the context of the discussion when replying to users. When the user says \"continue,\" please proceed with the continuation of the previous assistant's response.</s>\n",
-      "roles": [
-        "<|user|>\n",
-        "<|assistant|>\n"
-      ],
-      "intra_message_sep": "</s>\n",
-      "inter_message_sep": "</s>\n",
-      "stop_token_ids": [
-        2
-      ],
-      "stop": [
-        "</s>"
-      ]
-    }
+    "chat_template": "{% for item in messages %}{% if loop.first and item['role'] == 'system' %}{{ item['content'] + '\n' }}{% elif loop.first %}{{ '<|system|>\nYou are a creative super artificial intelligence assistant, possessing all the knowledge of humankind. Your name is csg-wukong, developed by OpenCSG. You need to understand and infer the true intentions of users based on the topics discussed in the chat history, and respond to user questions correctly as required. You enjoy responding to users with accurate and insightful answers. Please pay attention to the appropriate style and format when replying, try to avoid repetitive words and sentences, and keep your responses as concise and profound as possible. You carefully consider the context of the discussion when replying to users. When the user says \"continue,\" please proceed with the continuation of the previous assistant\\'s response.</s>\n' }}{% endif %}{% if item['role'] == 'user' %}{{ '<|user|>\n' + item['content'] + '</s>\n' }}{% elif item['role'] == 'assistant' %}{{ '<|assistant|>\n' + item['content'] + '</s>\n' }}{% endif %}{% endfor %}{% if add_generation_prompt %}{{ '<|assistant|>\n' }}{% endif %}",
+    "stop_token_ids": [
+      2
+    ],
+    "stop": [
+      "</s>"
+    ]
   }
 ]