ff

app.py

@@ -1,63 +1,136 @@
+import os
+import spaces
+
 import gradio as gr
-from huggingface_hub import InferenceClient
-
-"""
-For more information on `huggingface_hub` Inference API support, please check the docs: https://huggingface.co/docs/huggingface_hub/v0.22.2/en/guides/inference
-"""
-client = InferenceClient("HuggingFaceH4/zephyr-7b-beta")
-
-
-def respond(
-    message,
-    history: list[tuple[str, str]],
-    system_message,
-    max_tokens,
-    temperature,
-    top_p,
-):
-    messages = [{"role": "system", "content": system_message}]
-
-    for val in history:
-        if val[0]:
-            messages.append({"role": "user", "content": val[0]})
-        if val[1]:
-            messages.append({"role": "assistant", "content": val[1]})
-
-    messages.append({"role": "user", "content": message})
-
-    response = ""
-
-    for message in client.chat_completion(
-        messages,
-        max_tokens=max_tokens,
-        stream=True,
-        temperature=temperature,
-        top_p=top_p,
-    ):
-        token = message.choices[0].delta.content
-
-        response += token
-        yield response
-
-"""
-For information on how to customize the ChatInterface, peruse the gradio docs: https://www.gradio.app/docs/chatinterface
-"""
-demo = gr.ChatInterface(
-    respond,
-    additional_inputs=[
-        gr.Textbox(value="You are a friendly Chatbot.", label="System message"),
-        gr.Slider(minimum=1, maximum=2048, value=512, step=1, label="Max new tokens"),
-        gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature"),
-        gr.Slider(
-            minimum=0.1,
-            maximum=1.0,
-            value=0.95,
-            step=0.05,
-            label="Top-p (nucleus sampling)",
-        ),
-    ],
+import torch
+from colpali_engine.models.paligemma_colbert_architecture import ColPali
+from colpali_engine.trainer.retrieval_evaluator import CustomEvaluator
+from colpali_engine.utils.colpali_processing_utils import (
+    process_images,
+    process_queries,
 )
+from pdf2image import convert_from_path
+from PIL import Image
+from torch.utils.data import DataLoader
+from tqdm import tqdm
+from transformers import AutoProcessor
+
+# Load model
+model_name = "vidore/colpali-v1.2"
+token = os.environ.get("HF_TOKEN")
+model = ColPali.from_pretrained(
+    "vidore/colpaligemma-3b-pt-448-base", torch_dtype=torch.bfloat16, device_map="cuda", token = token).eval()
+
+model.load_adapter(model_name)
+model = model.eval()
+processor = AutoProcessor.from_pretrained(model_name, token = token)
+
+mock_image = Image.new("RGB", (448, 448), (255, 255, 255))
+
+
+@spaces.GPU
+def search(query: str, ds, images, k):
+
+    device = "cuda:0" if torch.cuda.is_available() else "cpu"
+    if device != model.device:
+        model.to(device)
+        
+    qs = []
+    with torch.no_grad():
+        batch_query = process_queries(processor, [query], mock_image)
+        batch_query = {k: v.to(device) for k, v in batch_query.items()}
+        embeddings_query = model(**batch_query)
+        qs.extend(list(torch.unbind(embeddings_query.to("cpu"))))
+
+    retriever_evaluator = CustomEvaluator(is_multi_vector=True)
+    scores = retriever_evaluator.evaluate(qs, ds)
+
+    top_k_indices = scores.argsort(axis=1)[0][-k:][::-1]
+
+    results = []
+    for idx in top_k_indices:
+        results.append((images[idx], f"Page {idx}"))
+
+    return results
+
+
+def index(files, ds):
+    print("Converting files")
+    images = convert_files(files)
+    print(f"Files converted with {len(images)} images.")
+    return index_gpu(images, ds)
+    
+
+
+def convert_files(files):
+    images = []
+    for f in files:
+        images.extend(convert_from_path(f, thread_count=4))
+
+    if len(images) >= 150:
+        raise gr.Error("The number of images in the dataset should be less than 150.")
+    return images
+
+
+@spaces.GPU
+def index_gpu(images, ds):
+    """Example script to run inference with ColPali"""
+    
+    # run inference - docs
+    dataloader = DataLoader(
+        images,
+        batch_size=4,
+        shuffle=False,
+        collate_fn=lambda x: process_images(processor, x),
+    )
+
+    
+    device = "cuda:0" if torch.cuda.is_available() else "cpu"
+    if device != model.device:
+        model.to(device)
+        
+          
+    for batch_doc in tqdm(dataloader):
+        with torch.no_grad():
+            batch_doc = {k: v.to(device) for k, v in batch_doc.items()}
+            embeddings_doc = model(**batch_doc)
+        ds.extend(list(torch.unbind(embeddings_doc.to("cpu"))))
+    return f"Uploaded and converted {len(images)} pages", ds, images
+
+
+def get_example():
+    return [[["climate_youth_magazine.pdf"], "How much tropical forest is cut annually ?"]]
+
+with gr.Blocks(theme=gr.themes.Soft()) as demo:
+    gr.Markdown("# ColPali: Efficient Document Retrieval with Vision Language Models 📚")
+
+    with gr.Row():
+        with gr.Column(scale=2):
+            gr.Markdown("## 1️⃣ Upload PDFs")
+            file = gr.File(file_types=["pdf"], file_count="multiple", label="Upload PDFs")
+
+            convert_button = gr.Button("🔄 Index documents")
+            message = gr.Textbox("Files not yet uploaded", label="Status")
+            embeds = gr.State(value=[])
+            imgs = gr.State(value=[])
+
+        with gr.Column(scale=3):
+            gr.Markdown("## 2️⃣ Search")
+            query = gr.Textbox(placeholder="Enter your query here", label="Query")
+            k = gr.Slider(minimum=1, maximum=10, step=1, label="Number of results", value=5)
+
+    # with gr.Row():
+    #    gr.Examples(
+    #        examples=get_example(),
+    #        inputs=[file, query],
+    #    )
+
+    # Define the actions
+    search_button = gr.Button("🔍 Search", variant="primary")
+    output_gallery = gr.Gallery(label="Retrieved Documents", height=600, show_label=True)
 
+    convert_button.click(index, inputs=[file, embeds], outputs=[message, embeds, imgs])
+    search_button.click(search, inputs=[query, embeds, imgs, k], outputs=[output_gallery])
 
 if __name__ == "__main__":
-    demo.launch()
+    demo.queue(max_size=10).launch(debug=True)

requirements.txt

@@ -1 +1,6 @@
-huggingface_hub==0.22.2
+huggingface_hub==0.22.2
+colpali-engine==0.2.0
+pdf2image
+GPUtil
+accelerate==0.30.1
+mteb>=1.12.22