Update modeling_text_encoder.py

modeling_text_encoder.py

@@ -43,17 +43,17 @@ class SD3TextEncoderWithMask(nn.Module):
         if self.text_encoder is None:
             self.text_encoder = CLIPTextModelWithProjection.from_pretrained(
                 os.path.join(self.model_path, 'text_encoder'), torch_dtype=self.torch_dtype
-            ).to(self.device_1)  # Move to GPU 0
+            ).to(self.device_0)  # Move to GPU 0
 
         if self.text_encoder_2 is None:
             self.text_encoder_2 = CLIPTextModelWithProjection.from_pretrained(
                 os.path.join(self.model_path, 'text_encoder_2'), torch_dtype=self.torch_dtype
-            ).to(self.device_1)  # Move to GPU 0
+            ).to(self.device_0)  # Move to GPU 0
 
         if self.text_encoder_3 is None:
             self.text_encoder_3 = T5EncoderModel.from_pretrained(
                 os.path.join(self.model_path, 'text_encoder_3'), torch_dtype=self.torch_dtype
-            ).to(self.device_1)  # Move to GPU 0
+            ).to(self.device_0)  # Move to GPU 0
 
     def _get_t5_prompt_embeds(
         self,