FEAT: [model] Support Qwen3-VL (#4112)

Co-authored-by: qinxuye <[email protected]>

Author: Jun-Howie

xinference/model/llm/llm_family.json

@@ -21617,5 +21617,174 @@
     ],
     "reasoning_start_tag": "<think>",
     "reasoning_end_tag": "</think>"
+  },
+  {
+    "version": 2,
+    "context_length": 262144,
+    "model_name": "Qwen3-VL-Instruct",
+    "model_lang": [
+      "en",
+      "zh"
+    ],
+    "model_ability": [
+      "chat",
+      "vision",
+      "tools"
+    ],
+    "model_description": "Meet Qwen3-VL — the most powerful vision-language model in the Qwen series to date.",
+    "model_specs": [
+      {
+        "model_format": "pytorch",
+        "model_size_in_billions": 235,
+        "activated_size_in_billions": 22,
+        "model_src": {
+          "huggingface": {
+            "quantizations": [
+              "none"
+            ],
+            "model_id": "Qwen/Qwen3-VL-235B-A22B-Instruct"
+          },
+          "modelscope": {
+            "quantizations": [
+              "none"
+            ],
+            "model_id": "Qwen/Qwen3-VL-235B-A22B-Instruct"
+          }
+        }
+      },
+      {
+        "model_format": "fp8",
+        "model_size_in_billions": 235,
+        "activated_size_in_billions": 22,
+        "model_src": {
+          "huggingface": {
+            "quantizations": [
+              "fp8"
+            ],
+            "model_id": "QuantTrio/Qwen3-VL-235B-A22B-Instruct-FP8"
+          },
+          "modelscope": {
+            "quantizations": [
+              "fp8"
+            ],
+            "model_id": "tclf90/Qwen3-VL-235B-A22B-Instruct-FP8"
+          }
+        }
+      },
+      {
+        "model_format": "awq",
+        "model_size_in_billions": 235,
+        "activated_size_in_billions": 22,
+        "model_src": {
+          "huggingface": {
+            "quantizations": [
+              "Int4"
+            ],
+            "model_id": "QuantTrio/Qwen3-VL-235B-A22B-Instruct-AWQ"
+          },
+          "modelscope": {
+            "quantizations": [
+              "Int4"
+            ],
+            "model_id": "tclf90/Qwen3-VL-235B-A22B-Instruct-AWQ"
+          }
+        }
+      }
+    ],
+    "chat_template": "{%- if tools %}\n    {{- '<|im_start|>system\\n' }}\n    {%- if messages[0].role == 'system' %}\n        {%- if messages[0].content is string %}\n            {{- messages[0].content }}\n        {%- else %}\n            {%- for content in messages[0].content %}\n                {%- if 'text' in content %}\n                    {{- content.text }}\n                {%- endif %}\n            {%- endfor %}\n        {%- endif %}\n        {{- '\\n\\n' }}\n    {%- endif %}\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' }}\n        {%- if messages[0].content is string %}\n            {{- messages[0].content }}\n        {%- else %}\n            {%- for content in messages[0].content %}\n                {%- if 'text' in content %}\n                    {{- content.text }}\n                {%- endif %}\n            {%- endfor %}\n        {%- endif %}\n        {{- '<|im_end|>\\n' }}\n    {%- endif %}\n{%- endif %}\n{%- set image_count = namespace(value=0) %}\n{%- set video_count = namespace(value=0) %}\n{%- for message in messages %}\n    {%- if message.role == \"user\" %}\n        {{- '<|im_start|>' + message.role + '\\n' }}\n        {%- if message.content is string %}\n            {{- message.content }}\n        {%- else %}\n            {%- for content in message.content %}\n                {%- if content.type == 'image' or 'image' in content or 'image_url' in content %}\n                    {%- set image_count.value = image_count.value + 1 %}\n                    {%- if add_vision_id %}Picture {{ image_count.value }}: {% endif -%}\n                    <|vision_start|><|image_pad|><|vision_end|>\n                {%- elif content.type == 'video' or 'video' in content %}\n                    {%- set video_count.value = video_count.value + 1 %}\n                    {%- if add_vision_id %}Video {{ video_count.value }}: {% endif -%}\n                    <|vision_start|><|video_pad|><|vision_end|>\n                {%- elif 'text' in content %}\n                    {{- content.text }}\n                {%- endif %}\n            {%- endfor %}\n        {%- endif %}\n        {{- '<|im_end|>\\n' }}\n    {%- elif message.role == \"assistant\" %}\n        {{- '<|im_start|>' + message.role + '\\n' }}\n        {%- if message.content is string %}\n            {{- message.content }}\n        {%- else %}\n            {%- for content_item in message.content %}\n                {%- if 'text' in content_item %}\n                    {{- content_item.text }}\n                {%- endif %}\n            {%- endfor %}\n        {%- endif %}\n        {%- if message.tool_calls %}\n            {%- for tool_call in message.tool_calls %}\n                {%- if (loop.first and message.content) or (not loop.first) %}\n                    {{- '\\n' }}\n                {%- endif %}\n                {%- if tool_call.function %}\n                    {%- set tool_call = tool_call.function %}\n                {%- endif %}\n                {{- '<tool_call>\\n{\"name\": \"' }}\n                {{- tool_call.name }}\n                {{- '\", \"arguments\": ' }}\n                {%- if tool_call.arguments is string %}\n                    {{- tool_call.arguments }}\n                {%- else %}\n                    {{- tool_call.arguments | tojson }}\n                {%- endif %}\n                {{- '}\\n</tool_call>' }}\n            {%- endfor %}\n        {%- endif %}\n        {{- '<|im_end|>\\n' }}\n    {%- elif message.role == \"tool\" %}\n        {%- if loop.first or (messages[loop.index0 - 1].role != \"tool\") %}\n            {{- '<|im_start|>user' }}\n        {%- endif %}\n        {{- '\\n<tool_response>\\n' }}\n        {%- if message.content is string %}\n            {{- message.content }}\n        {%- else %}\n            {%- for content in message.content %}\n                {%- if content.type == 'image' or 'image' in content or 'image_url' in content %}\n                    {%- set image_count.value = image_count.value + 1 %}\n                    {%- if add_vision_id %}Picture {{ image_count.value }}: {% endif -%}\n                    <|vision_start|><|image_pad|><|vision_end|>\n                {%- elif content.type == 'video' or 'video' in content %}\n                    {%- set video_count.value = video_count.value + 1 %}\n                    {%- if add_vision_id %}Video {{ video_count.value }}: {% endif -%}\n                    <|vision_start|><|video_pad|><|vision_end|>\n                {%- elif 'text' in content %}\n                    {{- content.text }}\n                {%- endif %}\n            {%- endfor %}\n        {%- endif %}\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",
+    "stop_token_ids": [
+      151643,
+      151645
+    ],
+    "stop": [
+      "<|endoftext|>",
+      "<|im_end|>"
+    ]
+  },
+  {
+    "version": 2,
+    "context_length": 262144,
+    "model_name": "Qwen3-VL-Thinking",
+    "model_lang": [
+      "en",
+      "zh"
+    ],
+    "model_ability": [
+      "chat",
+      "vision",
+      "reasoning",
+      "tools"
+    ],
+    "model_description": "Meet Qwen3-VL — the most powerful vision-language model in the Qwen series to date.",
+    "model_specs": [
+      {
+        "model_format": "pytorch",
+        "model_size_in_billions": 235,
+        "activated_size_in_billions": 22,
+        "model_src": {
+          "huggingface": {
+            "quantizations": [
+              "none"
+            ],
+            "model_id": "Qwen/Qwen3-VL-235B-A22B-Thinking"
+          },
+          "modelscope": {
+            "quantizations": [
+              "none"
+            ],
+            "model_id": "Qwen/Qwen3-VL-235B-A22B-Thinking"
+          }
+        }
+      },
+      {
+        "model_format": "fp8",
+        "model_size_in_billions": 235,
+        "activated_size_in_billions": 22,
+        "model_src": {
+          "huggingface": {
+            "quantizations": [
+              "fp8"
+            ],
+            "model_id": "QuantTrio/Qwen3-VL-235B-A22B-Thinking-FP8"
+          },
+          "modelscope": {
+            "quantizations": [
+              "fp8"
+            ],
+            "model_id": "tclf90/Qwen3-VL-235B-A22B-Thinking-FP8"
+          }
+        }
+      },
+      {
+        "model_format": "awq",
+        "model_size_in_billions": 235,
+        "activated_size_in_billions": 22,
+        "model_src": {
+          "huggingface": {
+            "quantizations": [
+              "Int4"
+            ],
+            "model_id": "QuantTrio/Qwen3-VL-235B-A22B-Thinking-AWQ"
+          },
+          "modelscope": {
+            "quantizations": [
+              "Int4"
+            ],
+            "model_id": "tclf90/Qwen3-VL-235B-A22B-Thinking-AWQ"
+          }
+        }
+      }
+    ],
+    "chat_template": "{%- set image_count = namespace(value=0) %}\n{%- set video_count = namespace(value=0) %}\n{%- macro render_content(content, do_vision_count) %}\n    {%- if content is string %}\n        {{- content }}\n    {%- else %}\n        {%- for item in content %}\n            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}\n                {%- if do_vision_count %}\n                    {%- set image_count.value = image_count.value + 1 %}\n                {%- endif %}\n                {%- if add_vision_id %}Picture {{ image_count.value }}: {% endif -%}\n                <|vision_start|><|image_pad|><|vision_end|>\n            {%- elif 'video' in item or item.type == 'video' %}\n                {%- if do_vision_count %}\n                    {%- set video_count.value = video_count.value + 1 %}\n                {%- endif %}\n                {%- if add_vision_id %}Video {{ video_count.value }}: {% endif -%}\n                <|vision_start|><|video_pad|><|vision_end|>\n            {%- elif 'text' in item %}\n                {{- item.text }}\n            {%- endif %}\n        {%- endfor %}\n    {%- endif %}\n{%- endmacro %}\n{%- if tools %}\n    {{- '<|im_start|>system\\n' }}\n    {%- if messages[0].role == 'system' %}\n        {{- render_content(messages[0].content, false) + '\\n\\n' }}\n    {%- endif %}\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' + render_content(messages[0].content, false) + '<|im_end|>\\n' }}\n    {%- endif %}\n{%- endif %}\n{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}\n{%- for message in messages[::-1] %}\n    {%- set index = (messages|length - 1) - loop.index0 %}\n    {%- if ns.multi_step_tool and message.role == \"user\" %}\n        {%- set content = render_content(message.content, false) %}\n        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}\n            {%- set ns.multi_step_tool = false %}\n            {%- set ns.last_query_index = index %}\n        {%- endif %}\n    {%- endif %}\n{%- endfor %}\n{%- for message in messages %}\n    {%- set content = render_content(message.content, True) %}\n    {%- if (message.role == \"user\") or (message.role == \"system\" and not loop.first) %}\n        {{- '<|im_start|>' + message.role + '\\n' + content + '<|im_end|>' + '\\n' }}\n    {%- elif message.role == \"assistant\" %}\n        {%- set reasoning_content = '' %}\n        {%- if message.reasoning_content is string %}\n            {%- set reasoning_content = message.reasoning_content %}\n        {%- else %}\n            {%- if '</think>' in content %}\n                {%- set reasoning_content = content.split('</think>')[0].rstrip('\\n').split('<think>')[-1].lstrip('\\n') %}\n                {%- set content = content.split('</think>')[-1].lstrip('\\n') %}\n            {%- endif %}\n        {%- endif %}\n        {%- if loop.index0 > ns.last_query_index %}\n            {%- if loop.last or (not loop.last and reasoning_content) %}\n                {{- '<|im_start|>' + message.role + '\\n<think>\\n' + reasoning_content.strip('\\n') + '\\n</think>\\n\\n' + content.lstrip('\\n') }}\n            {%- else %}\n                {{- '<|im_start|>' + message.role + '\\n' + content }}\n            {%- endif %}\n        {%- else %}\n            {{- '<|im_start|>' + message.role + '\\n' + content }}\n        {%- endif %}\n        {%- if message.tool_calls %}\n            {%- for tool_call in message.tool_calls %}\n                {%- if (loop.first and content) or (not loop.first) %}\n                    {{- '\\n' }}\n                {%- endif %}\n                {%- if tool_call.function %}\n                    {%- set tool_call = tool_call.function %}\n                {%- endif %}\n                {{- '<tool_call>\\n{\"name\": \"' }}\n                {{- tool_call.name }}\n                {{- '\", \"arguments\": ' }}\n                {%- if tool_call.arguments is string %}\n                    {{- tool_call.arguments }}\n                {%- else %}\n                    {{- tool_call.arguments | tojson }}\n                {%- endif %}\n                {{- '}\\n</tool_call>' }}\n            {%- endfor %}\n        {%- endif %}\n        {{- '<|im_end|>\\n' }}\n    {%- elif message.role == \"tool\" %}\n        {%- if loop.first or (messages[loop.index0 - 1].role != \"tool\") %}\n            {{- '<|im_start|>user' }}\n        {%- endif %}\n        {{- '\\n<tool_response>\\n' }}\n        {{- 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<think>\\n' }}\n{%- endif %}\n",
+    "stop_token_ids": [
+      151643,
+      151645
+    ],
+    "stop": [
+      "<|endoftext|>",
+      "<|im_end|>"
+    ],
+    "reasoning_start_tag": "<think>",
+    "reasoning_end_tag": "</think>"
   }
 ]

xinference/model/llm/transformers/multimodal/qwen2_vl.py

@@ -27,11 +27,19 @@
 
 
 @register_batching_multimodal_models(
-    "qwen2-vl-instruct", "qwen2.5-vl-instruct", "QvQ-72B-Preview"
+    "qwen2-vl-instruct",
+    "qwen2.5-vl-instruct",
+    "QvQ-72B-Preview",
+    "Qwen3-VL-Instruct",
+    "Qwen3-VL-Thinking",
 )
 @register_transformer
 @register_non_default_model(
-    "qwen2-vl-instruct", "qwen2.5-vl-instruct", "QvQ-72B-Preview"
+    "qwen2-vl-instruct",
+    "qwen2.5-vl-instruct",
+    "QvQ-72B-Preview",
+    "Qwen3-VL-Instruct",
+    "Qwen3-VL-Thinking",
 )
 class Qwen2VLChatModel(PytorchMultiModalModel):
     def _sanitize_model_config(
@@ -47,7 +55,7 @@ def _sanitize_model_config(
     def match_json(
         cls, model_family: "LLMFamilyV2", model_spec: "LLMSpecV1", quantization: str
     ) -> bool:
-        if model_spec.model_format not in ["pytorch", "gptq", "awq", "bnb"]:
+        if model_spec.model_format not in ["pytorch", "gptq", "awq", "bnb", "fp8"]:
             return False
         llm_family = model_family.model_family or model_family.model_name
         if "qwen2-vl-instruct".lower() in llm_family.lower():
@@ -56,6 +64,8 @@ def match_json(
             return True
         if "qvq-72b-preview".lower() in llm_family.lower():
             return True
+        if "qwen3-vl" in llm_family.lower():
+            return True
         return False
 
     def decide_device(self):
@@ -85,13 +95,19 @@ def load_multimodal_model(self):
         except ImportError:
             Qwen2_5_VLForConditionalGeneration = None
 
+        try:
+            from transformers import AutoModelForImageTextToText
+        except ImportError:
+            AutoModelForImageTextToText = None
+
         kwargs = self.apply_bnb_quantization()
         llm_family = self.model_family.model_family or self.model_family.model_name
-        model_cls = (
-            Qwen2_5_VLForConditionalGeneration
-            if "qwen2.5" in llm_family
-            else Qwen2VLForConditionalGeneration
-        )
+        if "qwen2.5" in llm_family:
+            model_cls = Qwen2_5_VLForConditionalGeneration
+        elif "qwen3" in llm_family:
+            model_cls = AutoModelForImageTextToText
+        else:
+            model_cls = Qwen2VLForConditionalGeneration
         if model_cls is None:
             raise ImportError("`transformers` version is too old, please upgrade it")
         device = "auto" if self._device == "cuda" else self._device

xinference/model/llm/utils.py

@@ -71,6 +71,8 @@
     "Qwen3-Thinking",
     "Qwen3-Instruct",
     "Qwen3-Coder",
+    "Qwen3-VL-Instruct",
+    "Qwen3-VL-Thinking",
 ]
 
 GLM4_TOOL_CALL_FAMILY = [

xinference/model/llm/vllm/core.py

@@ -289,6 +289,10 @@ class VLLMGenerateConfig(TypedDict, total=False):
 if VLLM_INSTALLED and VLLM_VERSION >= version.parse("0.10.2"):
     VLLM_SUPPORTED_CHAT_MODELS.append("seed-oss")
 
+if VLLM_INSTALLED and VLLM_VERSION > version.parse("0.10.2"):
+    VLLM_SUPPORTED_VISION_MODEL_LIST.append("Qwen3-VL-Instruct")
+    VLLM_SUPPORTED_VISION_MODEL_LIST.append("Qwen3-VL-Instruct")
+
 
 class VLLMModel(LLM):
     def __init__(

xinference/ui/web/ui/src/scenes/launch_model/data/data.js

@@ -44,6 +44,7 @@ export const additionalParameterTipList = {
     'pipeline_parallel_size',
     'enable_prefix_caching',
     'enable_chunked_prefill',
+    'enable_expert_parallel',
     'enforce_eager',
     'cpu_offload_gb',
     'disable_custom_all_reduce',