xinference 1.10.0__py3-none-any.whl → 1.11.0__py3-none-any.whl

xinference/model/image/model_spec.json

@@ -254,93 +254,8 @@
       "torch_dtype": "bfloat16"
     },
     "default_generate_config": {
-      "guidance_scale": 1.0,
-      "true_cfg_scale": 1.0
-    },
-    "virtualenv": {
-      "packages": [
-        "diffusers==0.35.1",
-        "peft>=0.17.0",
-        "#system_torch#",
-        "#system_numpy#"
-      ],
-      "no_build_isolation": true
-    }
-  },
-  {
-    "version": 2,
-    "model_name": "Qwen-Image-Edit",
-    "model_family": "stable_diffusion",
-    "model_ability": [
-      "image2image"
-    ],
-    "model_src": {
-      "huggingface": {
-        "model_id": "Qwen/Qwen-Image-Edit",
-        "model_revision": "0b71959872ea3bf4d106c578b7c480ebb133dba7",
-        "gguf_model_id": "QuantStack/Qwen-Image-Edit-GGUF",
-        "gguf_quantizations": [
-          "Q2_K",
-          "Q3_K_M",
-          "Q3_K_S",
-          "Q4_0",
-          "Q4_1",
-          "Q4_K_M",
-          "Q4_K_S",
-          "Q5_0",
-          "Q5_1",
-          "Q5_K_M",
-          "Q5_K_S",
-          "Q6_K",
-          "Q8_0"
-        ],
-        "gguf_model_file_name_template": "Qwen_Image_Edit-{quantization}.gguf",
-        "lightning_model_id": "lightx2v/Qwen-Image-Lightning",
-        "lightning_versions": [
-          "4steps-V1.0-bf16",
-          "4steps-V1.0",
-          "8steps-V1.0-bf16",
-          "8steps-V1.0"
-        ],
-        "lightning_model_file_name_template": "Qwen-Image-Edit-Lightning-{lightning_version}.safetensors"
-      },
-      "modelscope": {
-        "model_id": "Qwen/Qwen-Image-Edit",
-        "model_revision": "master",
-        "gguf_model_id": "QuantStack/Qwen-Image-Edit-GGUF",
-        "gguf_quantizations": [
-          "Q2_K",
-          "Q3_K_M",
-          "Q3_K_S",
-          "Q4_0",
-          "Q4_1",
-          "Q4_K_M",
-          "Q4_K_S",
-          "Q5_0",
-          "Q5_1",
-          "Q5_K_M",
-          "Q5_K_S",
-          "Q6_K",
-          "Q8_0"
-        ],
-        "gguf_model_file_name_template": "Qwen_Image_Edit-{quantization}.gguf",
-        "lightning_model_id": "lightx2v/Qwen-Image-Lightning",
-        "lightning_versions": [
-          "4steps-V1.0-bf16",
-          "4steps-V1.0",
-          "8steps-V1.0-bf16",
-          "8steps-V1.0"
-        ],
-        "lightning_model_file_name_template": "Qwen-Image-Edit-Lightning-{lightning_version}.safetensors"
-      }
-    },
-    "default_model_config": {
-      "quantize": true,
-      "quantize_text_encoder": "text_encoder",
-      "torch_dtype": "bfloat16"
-    },
-    "default_generate_config": {
-      "true_cfg_scale": 4.0
+      "guidance_scale": 1,
+      "true_cfg_scale": 1
     },
     "virtualenv": {
       "packages": [
@@ -497,7 +412,7 @@
       "transformer_quantization": "nf4"
     },
     "default_generate_config": {
-      "guidance_scale": 1.0,
+      "guidance_scale": 1,
       "num_inference_steps": 4
     },
     "model_src": {
@@ -575,7 +490,7 @@
       "text2image"
     ],
     "default_generate_config": {
-      "guidance_scale": 0.0,
+      "guidance_scale": 0,
       "num_inference_steps": 1
     },
     "model_src": {
@@ -597,7 +512,7 @@
       "text2image"
     ],
     "default_generate_config": {
-      "guidance_scale": 0.0,
+      "guidance_scale": 0,
       "num_inference_steps": 1
     },
     "model_src": {
@@ -842,5 +757,159 @@
         "model_revision": "master"
       }
     }
+  },
+  {
+    "version": 2,
+    "model_name": "Qwen-Image-Edit",
+    "model_family": "stable_diffusion",
+    "model_ability": [
+      "image2image"
+    ],
+    "model_src": {
+      "huggingface": {
+        "model_id": "Qwen/Qwen-Image-Edit",
+        "model_revision": "0b71959872ea3bf4d106c578b7c480ebb133dba7",
+        "gguf_model_id": "QuantStack/Qwen-Image-Edit-GGUF",
+        "gguf_quantizations": [
+          "Q2_K",
+          "Q3_K_M",
+          "Q3_K_S",
+          "Q4_0",
+          "Q4_1",
+          "Q4_K_M",
+          "Q4_K_S",
+          "Q5_0",
+          "Q5_1",
+          "Q5_K_M",
+          "Q5_K_S",
+          "Q6_K",
+          "Q8_0"
+        ],
+        "gguf_model_file_name_template": "Qwen_Image_Edit-{quantization}.gguf",
+        "lightning_model_id": "lightx2v/Qwen-Image-Lightning",
+        "lightning_versions": [
+          "4steps-V1.0-bf16",
+          "4steps-V1.0",
+          "8steps-V1.0-bf16",
+          "8steps-V1.0"
+        ],
+        "lightning_model_file_name_template": "Qwen-Image-Edit-Lightning-{lightning_version}.safetensors"
+      },
+      "modelscope": {
+        "model_id": "Qwen/Qwen-Image-Edit",
+        "model_revision": "master",
+        "gguf_model_id": "QuantStack/Qwen-Image-Edit-GGUF",
+        "gguf_quantizations": [
+          "Q2_K",
+          "Q3_K_M",
+          "Q3_K_S",
+          "Q4_0",
+          "Q4_1",
+          "Q4_K_M",
+          "Q4_K_S",
+          "Q5_0",
+          "Q5_1",
+          "Q5_K_M",
+          "Q5_K_S",
+          "Q6_K",
+          "Q8_0"
+        ],
+        "gguf_model_file_name_template": "Qwen_Image_Edit-{quantization}.gguf",
+        "lightning_model_id": "lightx2v/Qwen-Image-Lightning",
+        "lightning_versions": [
+          "4steps-V1.0-bf16",
+          "4steps-V1.0",
+          "8steps-V1.0-bf16",
+          "8steps-V1.0"
+        ],
+        "lightning_model_file_name_template": "Qwen-Image-Edit-Lightning-{lightning_version}.safetensors"
+      }
+    },
+    "default_model_config": {
+      "quantize": true,
+      "quantize_text_encoder": "text_encoder",
+      "torch_dtype": "bfloat16"
+    },
+    "default_generate_config": {
+      "true_cfg_scale": 4
+    },
+    "virtualenv": {
+      "packages": [
+        "diffusers==0.35.1",
+        "peft>=0.17.0",
+        "#system_torch#",
+        "#system_numpy#"
+      ],
+      "no_build_isolation": true
+    }
+  },
+  {
+    "version": 2,
+    "model_name": "Qwen-Image-Edit-2509",
+    "model_family": "stable_diffusion",
+    "model_ability": [
+      "image2image"
+    ],
+    "model_src": {
+      "huggingface": {
+        "model_id": "Qwen/Qwen-Image-Edit-2509",
+        "model_revision": "d3968ef930e841f4c73640fb8afa3b306a78167e",
+        "gguf_model_id": "QuantStack/Qwen-Image-Edit-2509-GGUF",
+        "gguf_quantizations": [
+          "Q2_K",
+          "Q3_K_M",
+          "Q3_K_S",
+          "Q4_0",
+          "Q4_1",
+          "Q4_K_M",
+          "Q4_K_S",
+          "Q5_0",
+          "Q5_1",
+          "Q5_K_M",
+          "Q5_K_S",
+          "Q6_K",
+          "Q8_0"
+        ],
+        "gguf_model_file_name_template": "Qwen-Image-Edit-2509-{quantization}.gguf"
+      },
+      "modelscope": {
+        "model_id": "Qwen/Qwen-Image-Edit-2509",
+        "model_revision": "master",
+        "gguf_model_id": "QuantStack/Qwen-Image-Edit-2509-GGUF",
+        "gguf_quantizations": [
+          "Q2_K",
+          "Q3_K_M",
+          "Q3_K_S",
+          "Q4_0",
+          "Q4_1",
+          "Q4_K_M",
+          "Q4_K_S",
+          "Q5_0",
+          "Q5_1",
+          "Q5_K_M",
+          "Q5_K_S",
+          "Q6_K",
+          "Q8_0"
+        ],
+        "gguf_model_file_name_template": "Qwen-Image-Edit-2509-{quantization}.gguf"
+      }
+    },
+    "default_model_config": {
+      "quantize": true,
+      "quantize_text_encoder": "text_encoder",
+      "torch_dtype": "bfloat16"
+    },
+    "default_generate_config": {
+      "true_cfg_scale": 4
+    },
+    "virtualenv": {
+      "packages": [
+        "git+https://github.com/huggingface/diffusers",
+        "peft>=0.17.0",
+        "#system_torch#",
+        "#system_numpy#"
+      ],
+      "no_build_isolation": true
+    }
   }
 ]

xinference/model/image/stable_diffusion/core.py

@@ -836,7 +836,7 @@ class DiffusionModel(SDAPIDiffusionModelMixin):
 
     def image_to_image(
         self,
-        image: PIL.Image,
+        image: Union[PIL.Image.Image, List[PIL.Image.Image]],
         prompt: Optional[Union[str, List[str]]] = None,
         n: int = 1,
         size: Optional[str] = None,
@@ -856,7 +856,10 @@ class DiffusionModel(SDAPIDiffusionModelMixin):
         if padding_image_to_multiple := kwargs.pop("padding_image_to_multiple", None):
             # Model like SD3 image to image requires image's height and width is times of 16
             # padding the image if specified
-            origin_x, origin_y = image.size
+            if isinstance(image, list):
+                origin_x, origin_y = image[0].size
+            else:
+                origin_x, origin_y = image.size
             kwargs["origin_size"] = (origin_x, origin_y)
             kwargs["is_padded"] = True
             image = self.pad_to_multiple(image, multiple=int(padding_image_to_multiple))
@@ -864,14 +867,20 @@ class DiffusionModel(SDAPIDiffusionModelMixin):
         if size:
             width, height = map(int, re.split(r"[^\d]+", size))
             if padding_image_to_multiple:
-                width, height = image.size
+                if isinstance(image, list):
+                    width, height = image[0].size
+                else:
+                    width, height = image.size
             kwargs["width"] = width
             kwargs["height"] = height
         else:
             # SD3 image2image cannot accept width and height
             allow_width_height = model_accept_param(["width", "height"], model)
             if allow_width_height:
-                kwargs["width"], kwargs["height"] = image.size
+                if isinstance(image, list):
+                    kwargs["width"], kwargs["height"] = image[0].size
+                else:
+                    kwargs["width"], kwargs["height"] = image.size
 
         # generate config for lightning
         self._gen_config_for_lightning(kwargs)

xinference/model/llm/__init__.py

@@ -159,6 +159,10 @@ def load_model_family_from_json(json_filename, target_families):
                     BUILTIN_LLM_PROMPT_STYLE[model_spec.model_name][
                         "reasoning_end_tag"
                     ] = model_spec.reasoning_end_tag
+                if model_spec.tool_parser:
+                    BUILTIN_LLM_PROMPT_STYLE[model_spec.model_name][
+                        "tool_parser"
+                    ] = model_spec.tool_parser
 
         # register model family
         if "chat" in model_spec.model_ability:
@@ -175,12 +179,12 @@ def _install():
     from .mlx.core import MLXChatModel, MLXModel, MLXVisionModel
     from .sglang.core import SGLANGChatModel, SGLANGModel, SGLANGVisionModel
     from .transformers.core import PytorchChatModel, PytorchModel
-    from .vllm.core import VLLMChatModel, VLLMModel, VLLMVisionModel
+    from .vllm.core import VLLMChatModel, VLLMModel, VLLMMultiModel
 
     # register llm classes.
     LLAMA_CLASSES.extend([XllamaCppModel])
     SGLANG_CLASSES.extend([SGLANGModel, SGLANGChatModel, SGLANGVisionModel])
-    VLLM_CLASSES.extend([VLLMModel, VLLMChatModel, VLLMVisionModel])
+    VLLM_CLASSES.extend([VLLMModel, VLLMChatModel, VLLMMultiModel])
     MLX_CLASSES.extend([MLXModel, MLXChatModel, MLXVisionModel])
     LMDEPLOY_CLASSES.extend([LMDeployModel, LMDeployChatModel])
     TRANSFORMERS_CLASSES.extend([PytorchChatModel, PytorchModel])