Update app.py

app.py

@@ -17,7 +17,6 @@ feature_extractor = ViTFeatureExtractor.from_pretrained(encoder_checkpoint)
 tokenizer = AutoTokenizer.from_pretrained(decoder_checkpoint)
 model = VisionEncoderDecoderModel.from_pretrained(model_checkpoint).to(device)
 
-
 def predict(image, max_length=64, num_beams=4):
   image = image.convert('RGB')
   image = feature_extractor(image, return_tensors="pt").pixel_values.to(device)
@@ -41,6 +40,7 @@ with gr.Blocks() as demo:
         </h1>   
         <h2 style="text-align: left; font-weight: 450; font-size: 1rem; margin-top: 2rem; margin-bottom: 1.5rem">
         In the field of large language models, the challenge of fine-tuning has long perplexed researchers. Microsoft, however, has unveiled an innovative solution called <b>Low-Rank Adaptation (LoRA)</b>. With the emergence of behemoth models like GPT-3 boasting billions of parameters, the cost of fine-tuning them for specific tasks or domains has become exorbitant.
+        <br>
         LoRA offers a groundbreaking approach by freezing the weights of pre-trained models and introducing trainable layers known as <b>rank-decomposition matrices in each transformer block</b>. This ingenious technique significantly reduces the number of trainable parameters and minimizes GPU memory requirements, as gradients no longer need to be computed for the majority of model weights.        
         <br>
         <br>
@@ -58,8 +58,7 @@ with gr.Blocks() as demo:
                 out = gr.outputs.Textbox(type="text",label="Captions")   
                 
     button.click(predict, inputs=[img], outputs=[out])
-
-    
+ 
     gr.Examples(
         examples=examples,
         inputs=img,