fix: fixed bugs

xinference/api/restful_api.py

@@ -841,12 +841,6 @@ async def launch_model(
                     detail="Invalid input. Allocated gpu must be a multiple of replica.",
                 )
 
-        if model_path is not None:
-            if not os.path.exists(model_path):
-                raise ValueError(
-                    f"Invalid input. `model_path`: {model_path} File or directory does not exist."
-                )
-
         if peft_model_config is not None:
             peft_model_config = PeftModelConfig.from_dict(peft_model_config)
         else:

xinference/core/worker.py

@@ -800,6 +800,11 @@ async def launch_builtin_model(
                 raise ValueError(
                     f"PEFT adaptors can only be applied to pytorch-like models"
                 )
+        if model_path is not None:
+            if not os.path.exists(model_path):
+                raise ValueError(
+                    f"Invalid input. `model_path`: {model_path} File or directory does not exist."
+                )
 
         assert model_uid not in self._model_uid_to_model
         self._check_model_is_valid(model_name, model_format)

xinference/model/audio/core.py

@@ -154,7 +154,7 @@ def create_audio_model_instance(
     **kwargs,
 ) -> Tuple[Union[WhisperModel, ChatTTSModel, CosyVoiceModel], AudioModelDescription]:
     model_spec = match_audio(model_name, download_hub)
-    if model_path is None:
+    if not model_path:
         model_path = cache(model_spec)
     model: Union[WhisperModel, ChatTTSModel, CosyVoiceModel]
     if model_spec.model_family == "whisper":

xinference/model/embedding/core.py

@@ -353,7 +353,7 @@ def create_embedding_model_instance(
     **kwargs,
 ) -> Tuple[EmbeddingModel, EmbeddingModelDescription]:
     model_spec = match_embedding(model_name, download_hub)
-    if model_path is None or model_path == "":
+    if not model_path:
         model_path = cache(model_spec)
 
     model = EmbeddingModel(model_uid, model_path, **kwargs)

xinference/model/flexible/core.py

@@ -218,7 +218,7 @@ def create_flexible_model_instance(
     **kwargs,
 ) -> Tuple[FlexibleModel, FlexibleModelDescription]:
     model_spec = match_flexible_model(model_name)
-    if model_path is None or model_path == "":
+    if not model_path:
         model_path = model_spec.model_uri
     launcher_name = model_spec.launcher
     launcher_args = model_spec.parser_args()

xinference/model/image/core.py

@@ -210,7 +210,7 @@ def create_image_model_instance(
         for name in controlnet:
             for cn_model_spec in model_spec.controlnet:
                 if cn_model_spec.model_name == name:
-                    if model_path is None or model_path == "":
+                    if not model_path:
                         model_path = cache(cn_model_spec)
                     controlnet_model_paths.append(model_path)
                     break
@@ -222,7 +222,7 @@ def create_image_model_instance(
             kwargs["controlnet"] = controlnet_model_paths[0]
         else:
             kwargs["controlnet"] = controlnet_model_paths
-    if model_path is None or model_path == "":
+    if not model_path:
         model_path = cache(model_spec)
     if peft_model_config is not None:
         lora_model = peft_model_config.peft_model

xinference/model/llm/core.py

@@ -222,7 +222,7 @@ def create_llm_model_instance(
     )
     logger.debug(f"Launching {model_uid} with {llm_cls.__name__}")
 
-    if model_path is None or model_path == "":
+    if not model_path:
         model_path = cache(llm_family, llm_spec, quantization)
 
     peft_model = peft_model_config.peft_model if peft_model_config else None

xinference/model/rerank/core.py

@@ -322,7 +322,7 @@ def create_rerank_model_instance(
                 f"Huggingface: {BUILTIN_RERANK_MODELS.keys()}"
                 f"ModelScope: {MODELSCOPE_RERANK_MODELS.keys()}"
             )
-    if model_path is None or model_path == "":
+    if not model_path:
         model_path = cache(model_spec)
     use_fp16 = kwargs.pop("use_fp16", False)
     model = RerankModel(

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

@@ -304,7 +304,7 @@ const ModelCard = ({
       worker_ip: workerIp.trim() === '' ? null : workerIp.trim(),
       gpu_idx: GPUIdx.trim() === '' ? null : handleGPUIdx(GPUIdx.trim()),
       download_hub: downloadHub === '' ? null : downloadHub,
-      model_path: modelPath === '' ? null : modelPath,
+      model_path: modelPath.trim() === '' ? null : modelPath.trim(),
     }
 
     let modelDataWithID_other = {
@@ -316,7 +316,7 @@ const ModelCard = ({
       worker_ip: workerIp.trim() === '' ? null : workerIp.trim(),
       gpu_idx: GPUIdx.trim() === '' ? null : handleGPUIdx(GPUIdx.trim()),
       download_hub: downloadHub === '' ? null : downloadHub,
-      model_path: modelPath === '' ? null : modelPath,
+      model_path: modelPath.trim() === '' ? null : modelPath.trim(),
     }
 
     if (nGPULayers >= 0) {