Spaces:
Sleeping
Sleeping
vk
commited on
Commit
·
5919b75
1
Parent(s):
c0118f4
first commit
Browse files- .gitattributes +1 -0
- .idea/.gitignore +3 -0
- app.py +64 -0
- requirements.txt +5 -0
- utils.py +41 -0
.gitattributes
CHANGED
@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
|
|
33 |
*.zip filter=lfs diff=lfs merge=lfs -text
|
34 |
*.zst filter=lfs diff=lfs merge=lfs -text
|
35 |
*tfevents* filter=lfs diff=lfs merge=lfs -text
|
|
|
|
33 |
*.zip filter=lfs diff=lfs merge=lfs -text
|
34 |
*.zst filter=lfs diff=lfs merge=lfs -text
|
35 |
*tfevents* filter=lfs diff=lfs merge=lfs -text
|
36 |
+
example/invoice1.png filter=lfs diff=lfs merge=lfs -text
|
.idea/.gitignore
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
# Default ignored files
|
2 |
+
/shelf/
|
3 |
+
/workspace.xml
|
app.py
ADDED
@@ -0,0 +1,64 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import gradio as gr
|
2 |
+
from peft import PeftModel, PeftConfig
|
3 |
+
from transformers import PaliGemmaForConditionalGeneration
|
4 |
+
import torch
|
5 |
+
from transformers import PaliGemmaProcessor
|
6 |
+
import PIL
|
7 |
+
from utils import parse_bbox_and_labels,display_boxes
|
8 |
+
|
9 |
+
|
10 |
+
|
11 |
+
|
12 |
+
|
13 |
+
|
14 |
+
|
15 |
+
def get_response(
|
16 |
+
image: PIL.Image.Image,
|
17 |
+
prompt: str,
|
18 |
+
max_new_tokens: str
|
19 |
+
) -> str:
|
20 |
+
raw_image = image.convert("RGB")
|
21 |
+
width, height = raw_image.size
|
22 |
+
|
23 |
+
inputs = processor(raw_image, prompt, return_tensors="pt").to(device)
|
24 |
+
with torch.inference_mode():
|
25 |
+
output = peft_model.generate(**inputs, max_new_tokens=int(max_new_tokens))
|
26 |
+
|
27 |
+
input_len = inputs["input_ids"].shape[-1]
|
28 |
+
output = processor.decode(output[0][input_len:], skip_special_tokens=True)
|
29 |
+
print(output)
|
30 |
+
if "loc" in output:
|
31 |
+
boxes, labels = parse_bbox_and_labels(output)
|
32 |
+
raw_image=display_boxes(raw_image, boxes, labels, target_size=(width, height))
|
33 |
+
|
34 |
+
return output,raw_image
|
35 |
+
|
36 |
+
|
37 |
+
if __name__ == "__main__":
|
38 |
+
|
39 |
+
device = torch.device("cpu")
|
40 |
+
# bnb_config = BitsAndBytesConfig(load_in_4bit=True, bnb_4bit_compute_dtype=torch.bfloat16) #for gpu
|
41 |
+
peft_model_id = "vk888/paligemma_vqav2"
|
42 |
+
model_id = "google/paligemma2-3b-pt-448"
|
43 |
+
config = PeftConfig.from_pretrained(peft_model_id)
|
44 |
+
base_model = PaliGemmaForConditionalGeneration.from_pretrained(config.base_model_name_or_path,
|
45 |
+
device_map=device) # , quantization_config=bnb_config)
|
46 |
+
peft_model = PeftModel.from_pretrained(base_model, peft_model_id)
|
47 |
+
processor = PaliGemmaProcessor.from_pretrained(model_id)
|
48 |
+
|
49 |
+
examples = [
|
50 |
+
["example/invoice1.png","<image>answer en what is the balance due ?\n", 80],
|
51 |
+
["example/invoice1.png","<image>detect signature\n", 80],
|
52 |
+
["example/invoice1.png","<image>answer en what is the rate cada of design ?\n", 80],
|
53 |
+
]
|
54 |
+
|
55 |
+
iface = gr.Interface(
|
56 |
+
cache_examples=False,
|
57 |
+
fn=get_response,
|
58 |
+
inputs=[gr.Image(type="pil"),gr.Textbox(placeholder="<image>answer en what is the balance due ?\n"),gr.Textbox(placeholder="80")],
|
59 |
+
examples=examples,
|
60 |
+
outputs=[gr.Textbox(), gr.Image(type="pil")],
|
61 |
+
title="DocVQA with Paligemma2 VLM",
|
62 |
+
description="DocVQA with Paligemma2 VLM"
|
63 |
+
)
|
64 |
+
iface.launch(share=True)
|
requirements.txt
ADDED
@@ -0,0 +1,5 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
--index-url https://download.pytorch.org/whl/cpu
|
2 |
+
torch
|
3 |
+
|
4 |
+
transformers==4.53.0.dev0
|
5 |
+
peft
|
utils.py
ADDED
@@ -0,0 +1,41 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import re
|
2 |
+
import numpy as np
|
3 |
+
from PIL import ImageDraw
|
4 |
+
|
5 |
+
def parse_bbox_and_labels(detokenized_output: str):
|
6 |
+
matches = re.finditer(
|
7 |
+
'<loc(?P<y0>\d\d\d\d)><loc(?P<x0>\d\d\d\d)><loc(?P<y1>\d\d\d\d)><loc(?P<x1>\d\d\d\d)>'
|
8 |
+
' (?P<label>.+?)( ;|$)',
|
9 |
+
detokenized_output,
|
10 |
+
)
|
11 |
+
labels, boxes = [], []
|
12 |
+
fmt = lambda x: float(x) / 1024.0
|
13 |
+
for m in matches:
|
14 |
+
d = m.groupdict()
|
15 |
+
boxes.append([fmt(d['y0']), fmt(d['x0']), fmt(d['y1']), fmt(d['x1'])])
|
16 |
+
labels.append(d['label'])
|
17 |
+
return np.array(boxes), np.array(labels)
|
18 |
+
|
19 |
+
def display_boxes(image, boxes, labels, target_size):
|
20 |
+
h, w = target_size
|
21 |
+
# fig, ax = plt.subplots()
|
22 |
+
# ax.imshow(image)
|
23 |
+
draw = ImageDraw.Draw(image)
|
24 |
+
for i in range(boxes.shape[0]):
|
25 |
+
y, x, y2, x2 = (boxes[i][0]*w,boxes[i][1]*h,boxes[i][2]*w,boxes[i][3]*h)
|
26 |
+
# width = x2 - x
|
27 |
+
# height = y2 - y
|
28 |
+
# Create a Rectangle patch
|
29 |
+
# rect = patches.Rectangle((x, y),
|
30 |
+
# width,
|
31 |
+
# height,
|
32 |
+
# linewidth=1,
|
33 |
+
# edgecolor='r',
|
34 |
+
# facecolor='none')
|
35 |
+
draw.rectangle((x,y,x2,y2) , outline="red", width=3)
|
36 |
+
# Add label
|
37 |
+
# plt.text(x, y, labels[i], color='red', fontsize=12)
|
38 |
+
# # Add the patch to the Axes
|
39 |
+
# ax.add_patch(rect)
|
40 |
+
# plt.show()
|
41 |
+
return image
|