更新 README.md 和 app.py，修改 HF_TOKEN 使用方式，增强模型加载的错误处理

README.md

@@ -12,7 +12,7 @@ Gradio Space using the official Z-Image pipeline (`Tongyi-MAI/Z-Image-Turbo`) wi
 1) Create a Space (Python) and select a GPU hardware type.  
 2) Add/clone this repo into the Space.  
 3) Manually add the LoRA file from https://civitai.com/models/2206377/zit-mystic-xxx to `lora/zit-mystic-xxx.safetensors` (or set `LORA_PATH`). Network fetch of Civitai is not handled in the Space.  
-4) (Recommended) Set `HF_TOKEN` in the Space secrets if the base model requires auth or to speed downloads.  
+4) If model download fails with a token error, set `HF_TOKEN` in the Space secrets (some repos require authentication).  
 5) (Optional) Toggle advanced envs below; then the Space will launch `app.py`. The header shows whether the LoRA was detected/loaded.
 
 ## Environment variables

app.py

@@ -135,7 +135,8 @@ def load_models() -> Tuple[ZImagePipeline, bool, str | None]:
     if pipe is not None:
         return pipe, lora_loaded, lora_error
 
-    use_auth_token = HF_TOKEN if HF_TOKEN else True
+    use_auth_token = HF_TOKEN if HF_TOKEN else None
+    hf_kwargs = {"use_auth_token": use_auth_token} if use_auth_token else {}
     print(f"Loading Z-Image from {MODEL_PATH}...")
 
     if not os.path.exists(MODEL_PATH):
@@ -143,15 +144,15 @@ def load_models() -> Tuple[ZImagePipeline, bool, str | None]:
             MODEL_PATH,
             subfolder="vae",
             torch_dtype=torch.bfloat16,
-            use_auth_token=use_auth_token,
+            **hf_kwargs,
         )
         text_encoder = AutoModelForCausalLM.from_pretrained(
             MODEL_PATH,
             subfolder="text_encoder",
             torch_dtype=torch.bfloat16,
-            use_auth_token=use_auth_token,
+            **hf_kwargs,
         ).eval()
-        tokenizer = AutoTokenizer.from_pretrained(MODEL_PATH, subfolder="tokenizer", use_auth_token=use_auth_token)
+        tokenizer = AutoTokenizer.from_pretrained(MODEL_PATH, subfolder="tokenizer", **hf_kwargs)
     else:
         vae = AutoencoderKL.from_pretrained(os.path.join(MODEL_PATH, "vae"), torch_dtype=torch.bfloat16)
         text_encoder = AutoModelForCausalLM.from_pretrained(
@@ -168,8 +169,8 @@ def load_models() -> Tuple[ZImagePipeline, bool, str | None]:
         transformer = ZImageTransformer2DModel.from_pretrained(
             MODEL_PATH,
             subfolder="transformer",
-            use_auth_token=use_auth_token,
             torch_dtype=torch.bfloat16,
+            **hf_kwargs,
         )
     else:
         transformer = ZImageTransformer2DModel.from_pretrained(
@@ -303,17 +304,20 @@ def warmup_model(pipeline: ZImagePipeline, resolutions: List[str]) -> None:
 
 
 def init_app() -> None:
-    ensure_models_loaded()
-    if ENABLE_WARMUP and pipe is not None:
-        ensure_on_gpu()
-        try:
-            all_resolutions: List[str] = []
-            for cat in RES_CHOICES.values():
-                all_resolutions.extend(cat)
-            warmup_model(pipe, all_resolutions)
-        finally:
-            if OFFLOAD_TO_CPU_AFTER_RUN:
-                offload_to_cpu()
+    try:
+        ensure_models_loaded()
+        if ENABLE_WARMUP and pipe is not None:
+            ensure_on_gpu()
+            try:
+                all_resolutions: List[str] = []
+                for cat in RES_CHOICES.values():
+                    all_resolutions.extend(cat)
+                warmup_model(pipe, all_resolutions)
+            finally:
+                if OFFLOAD_TO_CPU_AFTER_RUN:
+                    offload_to_cpu()
+    except Exception as exc:  # noqa: BLE001
+        print(f"Model init failed: {exc}")
 
 
 @spaces.GPU