Updated error handling

app.py

@@ -14,8 +14,8 @@ preprocessor_blip_large = AutoProcessor.from_pretrained("Salesforce/blip-image-c
 model_blip_large = BlipForConditionalGeneration.from_pretrained("Salesforce/blip-image-captioning-large")
 
 # Load the GIT coco model
-# preprocessor_git_large_coco = AutoProcessor.from_pretrained("microsoft/git-large-coco")
-# model_git_large_coco = AutoModelForCausalLM.from_pretrained("microsoft/git-large-coco")
+preprocessor_git_large_coco = AutoProcessor.from_pretrained("microsoft/git-large-coco")
+model_git_large_coco = AutoModelForCausalLM.from_pretrained("microsoft/git-large-coco")
 
 # Load the CLIP model
 model_oc_coca, _, transform_oc_coca = open_clip.create_model_and_transforms(
@@ -27,7 +27,7 @@ device = "cuda" if torch.cuda.is_available() else "cpu"
 # Transfer the models to the device
 model_blip_base.to(device)
 model_blip_large.to(device)
-# model_git_large_coco.to(device)
+model_git_large_coco.to(device)
 model_oc_coca.to(device)
 
 
@@ -103,11 +103,10 @@ def generate_captions_clip(
 	str
 		The generated caption.
 	"""
-	img = transform(image).unsqueeze(0).to(device)
+	im = transform(image).unsqueeze(0).to(device)
 	with torch.no_grad(), torch.cuda.amp.autocast():
-		generated = model.generate(img, seq_len=32, do_sample=True, temperature=0.9)
-	
-	generated_caption = model.decode(generated[0].detach()).split("<end_of_text>")[0].replace("<start_of_text>", "")
+		generated = model.generate(im, seq_len=20)
+	generated_caption = open_clip.decode(generated[0].detach()).split("<end_of_text>")[0].replace("<start_of_text>", "")
 	return generated_caption
 
 
@@ -129,17 +128,34 @@ def generate_captions(
 	str
 		The generated caption.
 	"""
+	caption_blip_base = ""
+	caption_blip_large = ""
+	caption_git_large_coco = ""
+	caption_oc_coca = ""
+
 	# Generate captions for the image using the Blip base model
-	caption_blip_base = generate_caption(preprocessor_blip_base, model_blip_base, image).strip()
+	try:
+		caption_blip_base = generate_caption(preprocessor_blip_base, model_blip_base, image).strip()
+	except Exception as e:
+		print(e)
 
 	# Generate captions for the image using the Blip large model
-	caption_blip_large = generate_caption(preprocessor_blip_large, model_blip_large, image).strip()
+	try:
+		caption_blip_large = generate_caption(preprocessor_blip_large, model_blip_large, image).strip()
+	except Exception as e:
+		print(e)
 	
 	# Generate captions for the image using the GIT coco model
-	# caption_git_large_coco = generate_caption(preprocessor_git_large_coco, model_git_large_coco, image).strip()
+	try:
+		caption_git_large_coco = generate_caption(preprocessor_git_large_coco, model_git_large_coco, image).strip()
+	except Exception as e:
+		print(e)
 	
 	# Generate captions for the image using the CLIP model
-	caption_oc_coca = generate_captions_clip(model_oc_coca, transform_oc_coca, image).strip()
+	try:
+		caption_oc_coca = generate_captions_clip(model_oc_coca, transform_oc_coca, image).strip()
+	except Exception as e:
+		print(e)
 
 	return caption_blip_base, caption_blip_large, caption_git_large_coco, caption_oc_coca
 
@@ -157,7 +173,7 @@ iface = gr.Interface(
 	outputs=[
 		gr.outputs.Textbox(label="Blip base"),
 		gr.outputs.Textbox(label="Blip large"),
-		# gr.outputs.Textbox(label="GIT large coco"),
+		gr.outputs.Textbox(label="GIT large coco"),
 		gr.outputs.Textbox(label="CLIP"),
 	],
 	title="Image Captioning",