Update app.py

app.py

@@ -21,7 +21,7 @@ orig_clip_model, orig_clip_processor = clip.load("ViT-B/32", device=device, jit=
 # Load the Unsplash dataset
 dataset = load_dataset("jamescalam/unsplash-25k-photos", split="train")  # all 25K images are in train split
 
-height = 256   # height for resizing images
+height = 512   # height for resizing images
 
 def predict(image, labels):
     inputs = processor(text=[f"a photo of {c}" for c in labels], images=image, return_tensors="pt", padding=True)
@@ -32,7 +32,7 @@ def predict(image, labels):
 
 
 def predict2(image, labels):
-    image = orig_clip_processor(img).unsqueeze(0).to(device)     
+    image = orig_clip_processor(image).unsqueeze(0).to(device)     
     text = clip.tokenize(labels).to(device)
     with torch.no_grad():
         image_features = orig_clip_model.encode_image(image)
@@ -157,6 +157,6 @@ with gr.Blocks() as demo:
             desc = gr.Textbox(show_label=False, placeholder="Enter description").style(container=False)
             search_btn = gr.Button("Find Images").style(full_width=False)
         gallery = gr.Gallery(show_label=False).style(grid=(2,2,3,5))
-        search_btn.click(search,inputs=desc, outputs=gallery)
+        search_btn.click(search,inputs=desc, outputs=gallery, postprocess=False)
 
 demo.launch()