modified: src/flux/util.py

app.py

@@ -341,7 +341,7 @@ class FluxEditor:
             self.model = self.model.to(self.device)
 
         self.model = self.model.to(self.device)
-        print("model has been moved to {self.device}")
+        print(f"model has been moved to {self.device}")
         
         #-----------------------------     1 Inverting current image     -------------------------------------#
         denoise_strategies = ['fireflow', 'rf', 'rf_solver', 'midpoint', 'rf_inversion', 'multi_turn_consistent']

src/flux/modules/conditioner.py

@@ -12,8 +12,8 @@ class HFEmbedder(nn.Module):
         self.output_key = "pooler_output" if self.is_clip else "last_hidden_state"
 
         if self.is_clip:
-            self.tokenizer: CLIPTokenizer = CLIPTokenizer.from_pretrained(version, max_length=max_length, truncation=True)
-            #self.tokenizer: CLIPTokenizer = CLIPTokenizer.from_pretrained("/home/user/app/models/tokenizer", max_length=max_length, truncation=True)
+            #self.tokenizer: CLIPTokenizer = CLIPTokenizer.from_pretrained(version, max_length=max_length, truncation=True)
+            self.tokenizer: CLIPTokenizer = CLIPTokenizer.from_pretrained("/home/user/app/models/tokenizer", max_length=max_length, truncation=True)
             self.hf_module: CLIPTextModel = CLIPTextModel.from_pretrained(version, **hf_kwargs)
 
             # --- DEBUG 信息 ---
@@ -75,7 +75,7 @@ class HFEmbedder(nn.Module):
         
         # 更严格的断言
         assert input_ids.shape == (len(text), self.max_length), f"Input IDs shape {input_ids.shape} does not match expected ({len(text)}, {self.max_length})"
-        print(f"Input IDs:\n{input_ids}")
+        #print(f"Input IDs:\n{input_ids}")
 
         # --- 关键调试：检查输入 ID 范围 ---
         min_id, max_id = input_ids.min().item(), input_ids.max().item()

src/flux/util.py

@@ -136,8 +136,8 @@ def load_t5(device: str | torch.device = "cuda", max_length: int = 77) -> HFEmbe
 
 
 def load_clip(device: str | torch.device = "cuda") -> HFEmbedder:
-    return HFEmbedder("openai/clip-vit-base-patch32", max_length=77, is_clip=True, torch_dtype=torch.bfloat16).to(device)
-    #return HFEmbedder("/home/user/app/models/text_encoder", max_length=77, is_clip=True, torch_dtype=torch.bfloat16).to(device)
+    #return HFEmbedder("openai/clip-vit-base-patch32", max_length=77, is_clip=True, torch_dtype=torch.bfloat16).to(device)
+    return HFEmbedder("/home/user/app/models/text_encoder", max_length=77, is_clip=True, torch_dtype=torch.bfloat16).to(device)
 
 
 def load_ae(name: str, device: str | torch.device = "cuda", hf_download: bool = True) -> AutoEncoder: