Update app.py

app.py

@@ -9,23 +9,17 @@ import torch
 import torchvision
 import subprocess
 
-# Run the commands from setup.sh to install poppler-utils
 def install_poppler():
     try:
         subprocess.run(["pdfinfo"], check=True, stdout=subprocess.PIPE, stderr=subprocess.PIPE)
     except FileNotFoundError:
         print("Poppler not found. Installing...")
-        # Run the setup commands
         subprocess.run("apt-get update", shell=True)
         subprocess.run("apt-get install -y poppler-utils", shell=True)
 
-# Call the Poppler installation check
 install_poppler()
-
-# Install flash-attn if not already installed
 subprocess.run('pip install flash-attn --no-build-isolation', env={'FLASH_ATTENTION_SKIP_CUDA_BUILD': "TRUE"}, shell=True)
 
-# Load the RAG Model and the Qwen2-VL-2B-Instruct model
 RAG = RAGMultiModalModel.from_pretrained("vidore/colpali")
 model = Qwen2VLForConditionalGeneration.from_pretrained("Qwen/Qwen2-VL-2B-Instruct",
                                                         trust_remote_code=True, torch_dtype=torch.bfloat16).cuda().eval()
@@ -33,24 +27,17 @@ processor = AutoProcessor.from_pretrained("Qwen/Qwen2-VL-2B-Instruct", trust_rem
 
 @spaces.GPU()
 def process_pdf_and_query(pdf_file, user_query):
-    # Convert the PDF to images
     images = convert_from_path(pdf_file.name)
     num_images = len(images)
-
-    # Indexing the PDF in RAG
     RAG.index(
         input_path=pdf_file.name,
-        index_name="image_index",  
+        index_name="image_index",
         store_collection_with_index=False,
         overwrite=True
     )
-
-    # Search the query in the RAG model
     results = RAG.search(user_query, k=1)
     if not results:
         return "No results found.", num_images
-
-    # Retrieve the page number and process image
     image_index = results[0]["page_num"] - 1
     messages = [
         {
@@ -64,8 +51,6 @@ def process_pdf_and_query(pdf_file, user_query):
             ],
         }
     ]
-
-    # Generate text with the Qwen model
     text = processor.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
     image_inputs, video_inputs = process_vision_info(messages)
     inputs = processor(
@@ -76,8 +61,6 @@ def process_pdf_and_query(pdf_file, user_query):
         return_tensors="pt",
     )
     inputs = inputs.to("cuda")
-    
-    # Generate the output response
     generated_ids = model.generate(**inputs, max_new_tokens=50)
     generated_ids_trimmed = [
         out_ids[len(in_ids):] for in_ids, out_ids in zip(inputs.input_ids, generated_ids)
@@ -85,29 +68,8 @@ def process_pdf_and_query(pdf_file, user_query):
     output_text = processor.batch_decode(
         generated_ids_trimmed, skip_special_tokens=True, clean_up_tokenization_spaces=False
     )
-
     return output_text[0], num_images
 
-
-pdf_input = gr.File(label="Upload PDF")  
-query_input = gr.Textbox(label="Enter your query", placeholder="Ask a question about the PDF") 
-output_text = gr.Textbox(label="Model Answer")  
-output_images = gr.Textbox(label="Number of Images in PDF")  
-
-
-explanation = """
-<div style="text-align: center; margin-bottom: 20px;">
-    <h2 style="font-weight: bold; font-size: 24px;">Multimodal RAG (Retrieval-Augmented Generation)</h2>
-    <p>
-        This application utilizes the ColPali model as a multimodal retriever, 
-        which retrieves relevant information from documents and generates answers 
-        using the Qwen/Qwen2-VL-2B-Instruct LLM (Large Language Model) 
-        via the Byaldi library, developed by Answer.ai.
-    </p>
-</div>
-"""
-
-
 footer = """
 <div style="text-align: center; margin-top: 20px;">
     <a href="https://www.linkedin.com/in/pejman-ebrahimi-4a60151a7/" target="_blank">LinkedIn</a> |
@@ -121,49 +83,30 @@ footer = """
 </div>
 """
 
+pdf_input = gr.File(label="Upload PDF")
+query_input = gr.Textbox(label="Enter your query", placeholder="Ask a question about the PDF")
+output_text = gr.Textbox(label="Model Answer")
+output_images = gr.Textbox(label="Number of Images in PDF")
+duplicate_button = gr.DuplicateButton(value="Duplicate Space for private use", elem_classes="duplicate-button", color="green")
+
+explanation_text = """
+<div style='text-align: center; font-size: 16px;'>
+    <h2 style='font-weight: bold;'>Multimodal RAG Overview</h2>
+    <p>This application utilizes a Multimodal RAG (Retrieve-and-Generate) approach, enabling users to query information from PDF documents 
+    by extracting relevant text and images. The ColPali model serves as a multimodal retriever, while the Byaldi library simplifies 
+    the integration of ColPali. The Qwen/Qwen2-VL-2B-Instruct LLM enhances the generation of responses based on the retrieved content.</p>
+</div>
+"""
 
 demo = gr.Interface(
     fn=process_pdf_and_query, 
-    inputs=[pdf_input, query_input],  
-    outputs=[output_text, output_images],  
-    title="<div style='text-align: center; font-weight: bold; font-size: 28px;'>Multimodal RAG with Image Query - By <a href='https://github.com/arad1367' style='color: blue;'>Pejman Ebrahimi</a></div>",
-    theme='freddyaboulton/dracula_revamped',
+    inputs=[pdf_input, query_input], 
+    outputs=[output_text, output_images], 
+    title="<div style='text-align: center; font-size: 24px; font-weight: bold;'>Multimodal RAG with Image Query</div>",
+    description=explanation_text,
+    theme='freddyaboulton/dracula_revamped'
 )
 
-
-with demo:
-    gr.HTML(explanation)  
-    gr.HTML(footer) 
-    gr.DuplicateButton(value="Duplicate Space for private use", elem_classes="duplicate-button")  
-
-    submit_btn = gr.Button("Submit", elem_classes="submit-button")  
-
-
-css = """
-<style>
-    .submit-button {
-        background-color: green;
-        color: white;
-        border: none;
-        border-radius: 5px;
-        padding: 10px 20px;
-        font-size: 16px;
-        cursor: pointer;
-        margin: 10px; /* Add some space between buttons */
-    }
-    .duplicate-button {
-        background-color: green;
-        color: white;
-        border: none;
-        border-radius: 5px;
-        padding: 10px 20px;
-        font-size: 16px;
-        cursor: pointer;
-        margin: 10px;  /* Add some space between buttons */
-    }
-</style>
-"""
-gr.HTML(css)
-
-# Launch the Gradio app
-demo.launch(debug=True)  
+demo.launch(debug=True)
+demo.append(duplicate_button)
+demo.append(gr.HTML(footer))