Update app.py

app.py

@@ -272,17 +272,6 @@ def infer(ref_style_file, style_description, caption):
         # Reset the state after inference, regardless of success or failure
         reset_inference_state()
 
-def transform(tensor):
-    """
-    Define the necessary transformations for the image.
-    """
-    to_pil = T.ToPILImage()
-    return T.Compose([
-        T.Resize((256, 256)),  # Example resize, adjust as needed
-        T.ToTensor(),
-        T.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225]),
-    ])(to_pil(tensor))
-
 def infer_compo(style_description, ref_style_file, caption, ref_sub_file):
     global models_rbm, models_b
     try:
@@ -324,7 +313,8 @@ def infer_compo(style_description, ref_style_file, caption, ref_sub_file):
         
         # Convert tensor to PIL Image before passing to sam_model.predict
         x0_preview_pil = T.ToPILImage()(x0_preview[0])
-        sam_mask, boxes, phrases, logits = sam_model.predict(transform(x0_preview_pil), sam_prompt)
+        x0_preview_tensor = T.ToTensor()(x0_preview_pil)  # Convert PIL Image back to tensor
+        sam_mask, boxes, phrases, logits = sam_model.predict(x0_preview_pil, sam_prompt)
         sam_mask = sam_mask.detach().unsqueeze(dim=0).to(device)
         
         conditions = core.get_conditions(batch, models_rbm, extras, is_eval=True, is_unconditional=False, eval_image_embeds=True, eval_subject_style=True, eval_csd=False)