Use vit

app.py

@@ -6,17 +6,17 @@ torch.hub.download_url_to_file('http://images.cocodataset.org/val2017/0000000397
 torch.hub.download_url_to_file('https://huggingface.co/datasets/nielsr/textcaps-sample/resolve/main/stop_sign.png', 'stop_sign.png')
 torch.hub.download_url_to_file('https://cdn.openai.com/dall-e-2/demos/text2im/astronaut/horse/photo/0.jpg', 'astronaut.jpg')
 
-git_processor_base = AutoProcessor.from_pretrained("microsoft/git-base-coco")
-git_model_base = AutoModelForCausalLM.from_pretrained("microsoft/git-base-coco")
+# git_processor_base = AutoProcessor.from_pretrained("microsoft/git-base-coco")
+# git_model_base = AutoModelForCausalLM.from_pretrained("microsoft/git-base-coco")
 
-git_processor_large = AutoProcessor.from_pretrained("microsoft/git-large-coco")
-git_model_large = AutoModelForCausalLM.from_pretrained("microsoft/git-large-coco")
+# git_processor_large = AutoProcessor.from_pretrained("microsoft/git-large-coco")
+# git_model_large = AutoModelForCausalLM.from_pretrained("microsoft/git-large-coco")
 
-blip_processor_base = AutoProcessor.from_pretrained("Salesforce/blip-image-captioning-base")
-blip_model_base = BlipForConditionalGeneration.from_pretrained("Salesforce/blip-image-captioning-base")
+# blip_processor_base = AutoProcessor.from_pretrained("Salesforce/blip-image-captioning-base")
+# blip_model_base = BlipForConditionalGeneration.from_pretrained("Salesforce/blip-image-captioning-base")
 
-blip_processor_large = AutoProcessor.from_pretrained("Salesforce/blip-image-captioning-large")
-blip_model_large = BlipForConditionalGeneration.from_pretrained("Salesforce/blip-image-captioning-large")
+# blip_processor_large = AutoProcessor.from_pretrained("Salesforce/blip-image-captioning-large")
+# blip_model_large = BlipForConditionalGeneration.from_pretrained("Salesforce/blip-image-captioning-large")
 
 vitgpt_processor = AutoImageProcessor.from_pretrained("nlpconnect/vit-gpt2-image-captioning")
 vitgpt_model = VisionEncoderDecoderModel.from_pretrained("nlpconnect/vit-gpt2-image-captioning")
@@ -24,10 +24,10 @@ vitgpt_tokenizer = AutoTokenizer.from_pretrained("nlpconnect/vit-gpt2-image-capt
 
 device = "cuda" if torch.cuda.is_available() else "cpu"
 
-git_model_base.to(device)
-blip_model_base.to(device)
-git_model_large.to(device)
-blip_model_large.to(device)
+# git_model_base.to(device)
+# blip_model_base.to(device)
+# git_model_large.to(device)
+# blip_model_large.to(device)
 vitgpt_model.to(device)
 
 def generate_caption(processor, model, image, tokenizer=None):
@@ -44,21 +44,21 @@ def generate_caption(processor, model, image, tokenizer=None):
 
 
 def generate_captions(image):
-    caption_git_base = generate_caption(git_processor_base, git_model_base, image)
+    # caption_git_base = generate_caption(git_processor_base, git_model_base, image)
 
-    caption_git_large = generate_caption(git_processor_large, git_model_large, image)
+    # caption_git_large = generate_caption(git_processor_large, git_model_large, image)
 
-    caption_blip_base = generate_caption(blip_processor_base, blip_model_base, image)
+    # caption_blip_base = generate_caption(blip_processor_base, blip_model_base, image)
 
-    caption_blip_large = generate_caption(blip_processor_large, blip_model_large, image)
+    # caption_blip_large = generate_caption(blip_processor_large, blip_model_large, image)
 
     caption_vitgpt = generate_caption(vitgpt_processor, vitgpt_model, image, vitgpt_tokenizer)
 
-    return caption_git_base, caption_git_large, caption_blip_base, caption_blip_large, caption_vitgpt
+    return caption_vitgpt
 
    
 examples = [["cats.jpg"], ["stop_sign.png"], ["astronaut.jpg"]]
-outputs = [gr.outputs.Textbox(label="Caption generated by GIT-base"), gr.outputs.Textbox(label="Caption generated by GIT-large"), gr.outputs.Textbox(label="Caption generated by BLIP-base"), gr.outputs.Textbox(label="Caption generated by BLIP-large"), gr.outputs.Textbox(label="Caption generated by ViT+GPT-2")] 
+outputs = [gr.outputs.Textbox(label="Caption generated by ViT+GPT-2")] 
 
 title = "Interactive demo: comparing image captioning models"
 description = "Gradio Demo to compare GIT, BLIP and ViT+GPT2, 3 state-of-the-art vision+language models. To use it, simply upload your image and click 'submit', or click one of the examples to load them. Read more at the links below."