Create app.py
Browse files
app.py
ADDED
@@ -0,0 +1,74 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
from PIL import Image
|
2 |
+
import gradio as gr
|
3 |
+
import torch
|
4 |
+
import requests
|
5 |
+
from models.zhclip import ZhCLIPProcessor, ZhCLIPModel # From https://www.github.com/yue-gang/ZH-CLIP
|
6 |
+
|
7 |
+
version = 'nlpcver/zh-clip-vit-roberta-large-patch14'
|
8 |
+
model = ZhCLIPModel.from_pretrained(version)
|
9 |
+
processor = ZhCLIPProcessor.from_pretrained(version)
|
10 |
+
|
11 |
+
|
12 |
+
|
13 |
+
def get_result(image,text):
|
14 |
+
inputs = processor(text=[text], images=image, return_tensors="pt", padding=True)
|
15 |
+
|
16 |
+
outputs = model(**inputs)
|
17 |
+
image_features = outputs.image_features
|
18 |
+
text_features = outputs.text_features
|
19 |
+
text_probs = (image_features @ text_features.T).softmax(dim=-1)
|
20 |
+
return text_probs
|
21 |
+
with gr.Blocks(
|
22 |
+
css="""
|
23 |
+
.message.svelte-w6rprc.svelte-w6rprc.svelte-w6rprc {font-size: 20px; margin-top: 20px}
|
24 |
+
#component-21 > div.wrap.svelte-w6rprc {height: 600px;}
|
25 |
+
"""
|
26 |
+
) as iface:
|
27 |
+
state = gr.State([])
|
28 |
+
|
29 |
+
with gr.Row():
|
30 |
+
with gr.Column(scale=1):
|
31 |
+
image_input = gr.Image(type="pil",label="VQA Image Input")
|
32 |
+
with gr.Row():
|
33 |
+
with gr.Column(scale=1):
|
34 |
+
chat_input = gr.Textbox(lines=1, label="VQA Question Input")
|
35 |
+
with gr.Row():
|
36 |
+
clear_button = gr.Button(value="Clear", interactive=True,width=30)
|
37 |
+
submit_button = gr.Button(
|
38 |
+
value="Submit", interactive=True, variant="primary"
|
39 |
+
)
|
40 |
+
'''
|
41 |
+
cap_submit_button = gr.Button(
|
42 |
+
value="Submit_CAP", interactive=True, variant="primary"
|
43 |
+
)
|
44 |
+
gpt3_submit_button = gr.Button(
|
45 |
+
value="Submit_GPT3", interactive=True, variant="primary"
|
46 |
+
)
|
47 |
+
'''
|
48 |
+
with gr.Column():
|
49 |
+
caption_output = gr.Textbox(lines=0, label="ITM")
|
50 |
+
|
51 |
+
chat_input.submit(
|
52 |
+
get_result,
|
53 |
+
[
|
54 |
+
image_input,
|
55 |
+
chat_input,
|
56 |
+
],
|
57 |
+
[ caption_output],
|
58 |
+
)
|
59 |
+
clear_button.click(
|
60 |
+
lambda: ("", [],"","",""),
|
61 |
+
[],
|
62 |
+
[chat_input, state,caption_output],
|
63 |
+
queue=False,
|
64 |
+
)
|
65 |
+
submit_button.click(
|
66 |
+
get_result,
|
67 |
+
[
|
68 |
+
image_input,
|
69 |
+
chat_input,
|
70 |
+
],
|
71 |
+
[caption_output],
|
72 |
+
)
|
73 |
+
iface.queue(concurrency_count=1, api_open=False, max_size=10)
|
74 |
+
iface.launch(enable_queue=True)
|