Update app.py

app.py

@@ -1,5 +1,5 @@
 import gradio as gr
-from transformers import AutoProcessor, AutoTokenizer, AutoImageProcessor, AutoModelForCausalLM, BlipForConditionalGeneration, VisionEncoderDecoderModel
+from transformers import AutoProcessor, AutoTokenizer, AutoImageProcessor, AutoModelForCausalLM, BlipForConditionalGeneration, Blip2ForConditionalGeneration, VisionEncoderDecoderModel
 import torch
 import open_clip
 
@@ -24,9 +24,12 @@ git_model_large_textcaps = AutoModelForCausalLM.from_pretrained("microsoft/git-l
 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")
-vitgpt_tokenizer = AutoTokenizer.from_pretrained("nlpconnect/vit-gpt2-image-captioning")
+blip2_processor = AutoProcessor.from_pretrained("Salesforce/blip2-opt-2.7b")
+blip2_model = Blip2ForConditionalGeneration.from_pretrained("Salesforce/blip2-opt-2.7b")
+
+# vitgpt_processor = AutoImageProcessor.from_pretrained("nlpconnect/vit-gpt2-image-captioning")
+# vitgpt_model = VisionEncoderDecoderModel.from_pretrained("nlpconnect/vit-gpt2-image-captioning")
+# vitgpt_tokenizer = AutoTokenizer.from_pretrained("nlpconnect/vit-gpt2-image-captioning")
 
 coca_model, _, coca_transform = open_clip.create_model_and_transforms(
   model_name="coca_ViT-L-14",
@@ -40,8 +43,9 @@ device = "cuda" if torch.cuda.is_available() else "cpu"
 git_model_large_coco.to(device)
 git_model_large_textcaps.to(device)
 blip_model_large.to(device)
-vitgpt_model.to(device)
+# vitgpt_model.to(device)
 coca_model.to(device)
+blip2_model.to(device)
 
 def generate_caption(processor, model, image, tokenizer=None):
     inputs = processor(images=image, return_tensors="pt").to(device)
@@ -74,18 +78,20 @@ def generate_captions(image):
 
     caption_blip_large = generate_caption(blip_processor_large, blip_model_large, image)
 
-    caption_vitgpt = generate_caption(vitgpt_processor, vitgpt_model, image, vitgpt_tokenizer)
+    # caption_vitgpt = generate_caption(vitgpt_processor, vitgpt_model, image, vitgpt_tokenizer)
 
     caption_coca = generate_caption_coca(coca_model, coca_transform, image)
 
-    return caption_git_large_coco, caption_git_large_textcaps, caption_blip_large, caption_vitgpt, caption_coca
+    caption_blip2 = generate_caption(blip2_processor, blip2_model, image)
+
+    return caption_git_large_coco, caption_git_large_textcaps, caption_blip_large, caption_coca, caption_blip2
 
    
 examples = [["cats.jpg"], ["stop_sign.png"], ["astronaut.jpg"]]
-outputs = [gr.outputs.Textbox(label="Caption generated by GIT-large fine-tuned on COCO"), gr.outputs.Textbox(label="Caption generated by GIT-large fine-tuned on TextCaps"), gr.outputs.Textbox(label="Caption generated by BLIP-large"), gr.outputs.Textbox(label="Caption generated by ViT+GPT-2"), gr.outputs.Textbox(label="Caption generated by CoCa")] 
+outputs = [gr.outputs.Textbox(label="Caption generated by GIT-large fine-tuned on COCO"), gr.outputs.Textbox(label="Caption generated by GIT-large fine-tuned on TextCaps"), gr.outputs.Textbox(label="Caption generated by BLIP-large"), gr.outputs.Textbox(label="Caption generated by CoCa"), gr.outputs.Textbox(label="Caption generated by BLIP-2 OPT 2.7b")] 
 
 title = "Interactive demo: comparing image captioning models"
-description = "Gradio Demo to compare GIT, BLIP, ViT+GPT2 and CoCa, 4 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."
+description = "Gradio Demo to compare GIT, BLIP, CoCa, and BLIP-2, 4 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."
 article = "<p style='text-align: center'><a href='https://huggingface.co/docs/transformers/main/model_doc/blip' target='_blank'>BLIP docs</a> | <a href='https://huggingface.co/docs/transformers/main/model_doc/git' target='_blank'>GIT docs</a></p>"
 
 interface = gr.Interface(fn=generate_captions,