Spaces:
Sleeping
Sleeping
File size: 7,930 Bytes
8001e7f e3f404d 8001e7f e3f404d 8001e7f 8573be6 79fd59c bcc154b 79fd59c b7dad17 79fd59c 8573be6 79fd59c 8573be6 c34d360 adafc6a a39ac5a adafc6a f0726db 8573be6 e2b8125 8573be6 79fd59c 8573be6 79fd59c 8573be6 8001e7f 9f28ec7 8001e7f a791fbb a95df8d 28ef079 8001e7f a95df8d 8001e7f a95df8d 8001e7f 79fd59c 8001e7f affda77 e8e541b affda77 8001e7f 8fb0879 fc00b82 8001e7f caf5c98 8001e7f caf5c98 8001e7f 79fd59c 8001e7f caf5c98 fc00b82 8001e7f caf5c98 8001e7f e3f404d 8001e7f dc5eecc e3f404d cef7a44 affda77 dc5eecc cef7a44 e3f404d 8001e7f |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 |
import os
import spaces
import gradio as gr
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 Qwen2VLForConditionalGeneration, AutoTokenizer, AutoProcessor
from qwen_vl_utils import process_vision_info
import re
import time
from PIL import Image
import torch
import subprocess
# subprocess.run('pip install flash-attn --no-build-isolation', env={'FLASH_ATTENTION_SKIP_CUDA_BUILD': "TRUE"}, shell=True)
@spaces.GPU
def model_inference(
images, text,
):
print(type(images))
print(images[0])
images = Image.open(images[0][0])
print(images)
print(type(images))
# model = Qwen2VLForConditionalGeneration.from_pretrained(
# "Qwen/Qwen2-VL-7B-Instruct", torch_dtype="auto", device_map="auto"
# )
#We recommend enabling flash_attention_2 for better acceleration and memory saving, especially in multi-image and video scenarios.
model = Qwen2VLForConditionalGeneration.from_pretrained("Qwen/Qwen2-VL-2B-Instruct", trust_remote_code=True, torch_dtype="auto").cuda().eval()
# default processer
processor = AutoProcessor.from_pretrained("Qwen/Qwen2-VL-7B-Instruct")
# The default range for the number of visual tokens per image in the model is 4-16384. You can set min_pixels and max_pixels according to your needs, such as a token count range of 256-1280, to balance speed and memory usage.
# min_pixels = 256*28*28
# max_pixels = 1280*28*28
# processor = AutoProcessor.from_pretrained("Qwen/Qwen2-VL-7B-Instruct", min_pixels=min_pixels, max_pixels=max_pixels)
messages = [
{
"role": "user",
"content": [
{
"type": "image",
"image": images,
},
{"type": "text", "text": text},
],
}
]
# Preparation for inference
text = processor.apply_chat_template(
messages, tokenize=False, add_generation_prompt=True
)
image_inputs, video_inputs = process_vision_info(messages)
inputs = processor(
text=[text],
images=image_inputs,
videos=video_inputs,
padding=True,
return_tensors="pt",
)
inputs = inputs.to("cuda")
# Inference: Generation of the output
generated_ids = model.generate(**inputs, max_new_tokens=128)
generated_ids_trimmed = [
out_ids[len(in_ids) :] for in_ids, out_ids in zip(inputs.input_ids, generated_ids)
]
output_text = processor.batch_decode(
generated_ids_trimmed, skip_special_tokens=True, clean_up_tokenization_spaces=False
)
return output_text[0]
@spaces.GPU
def search(query: str, ds, images, k):
# Load colpali 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))
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}"
del model
del processor
print("done")
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"""
# Load colpali 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))
# 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"))))
del model
del processor
print("done")
return f"Uploaded and converted {len(images)} pages", ds, images
def get_example():
return [
[["RAPPORT_DEVELOPPEMENT_DURABLE_2019.pdf"], "Quels sont les 4 axes majeurs des achats?"],
[["RAPPORT_DEVELOPPEMENT_DURABLE_2019.pdf"], "Quelles sont les actions entreprise en Afrique du Sud?"],
[["RAPPORT_DEVELOPPEMENT_DURABLE_2019.pdf"], "fais moi un tableau de la répartition homme femme"],
]
with gr.Blocks(theme=gr.themes.Soft()) as demo:
gr.Markdown("# ColPali + Idefics3: 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")
message = gr.Textbox("Files not yet uploaded", label="Status")
convert_button = gr.Button("🔄 Index documents")
embeds = gr.State(value=[])
imgs = gr.State(value=[])
img_chunk = 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)
search_button = gr.Button("🔍 Search", variant="primary")
with gr.Row():
gr.Examples(
examples=get_example(),
inputs=[file, query],
)
# Define the actions
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])
answer_button = gr.Button("Answer", variant="primary")
output = gr.Markdown(label="Output")
answer_button.click(model_inference, inputs=[output_gallery, query], outputs=output)
if __name__ == "__main__":
demo.queue(max_size=10).launch(debug=True) |