Spaces:
Running
on
L40S
Running
on
L40S
Upload 5 files
Browse files- Dockerfile +31 -0
- app.py +161 -0
- requirements.txt +2 -0
- run.sh +44 -0
- utils/gradio_helpers.py +469 -0
Dockerfile
ADDED
@@ -0,0 +1,31 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
FROM r8.im/fofr/expression-editor@sha256:bf913bc90e1c44ba288ba3942a538693b72e8cc7df576f3beebe56adc0a92b86
|
2 |
+
RUN apt-get update && apt-get install -y netcat jq
|
3 |
+
|
4 |
+
RUN useradd -m -u 1000 user
|
5 |
+
RUN chown -R user:user / || true
|
6 |
+
RUN chown -R user:user /src/
|
7 |
+
RUN chown -R user:user /root/
|
8 |
+
RUN chown -R user:user /var/
|
9 |
+
USER user
|
10 |
+
ENV HOME=/home/user \
|
11 |
+
PATH=/home/user/.local/bin:$PATH \
|
12 |
+
PYTHONPATH=$HOME/app \
|
13 |
+
PYTHONUNBUFFERED=1 \
|
14 |
+
GRADIO_ALLOW_FLAGGING=never \
|
15 |
+
GRADIO_NUM_PORTS=1 \
|
16 |
+
GRADIO_SERVER_NAME=0.0.0.0 \
|
17 |
+
GRADIO_THEME=huggingface \
|
18 |
+
SYSTEM=spaces
|
19 |
+
|
20 |
+
WORKDIR $HOME/app
|
21 |
+
COPY ./requirements.txt /code/requirements.txt
|
22 |
+
|
23 |
+
# create virtual env for Gradio app
|
24 |
+
RUN python -m venv $HOME/.venv && \
|
25 |
+
. $HOME/.venv/bin/activate && \
|
26 |
+
pip install --no-cache-dir --upgrade pip && \
|
27 |
+
pip install --no-cache-dir -r /code/requirements.txt
|
28 |
+
|
29 |
+
COPY --chown=user . $HOME/app
|
30 |
+
RUN chmod +x $HOME/app/run.sh
|
31 |
+
CMD ["bash", "-c", "$HOME/app/run.sh"]
|
app.py
ADDED
@@ -0,0 +1,161 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import gradio as gr
|
2 |
+
from urllib.parse import urlparse
|
3 |
+
import requests
|
4 |
+
import time
|
5 |
+
import os
|
6 |
+
|
7 |
+
from utils.gradio_helpers import parse_outputs, process_outputs
|
8 |
+
|
9 |
+
inputs = []
|
10 |
+
inputs.append(gr.Image(
|
11 |
+
label="Image", type="filepath"
|
12 |
+
))
|
13 |
+
|
14 |
+
inputs.append(gr.Slider(
|
15 |
+
label="Rotate Pitch", info='''Rotation pitch: Adjusts the up and down tilt of the face''', value=0,
|
16 |
+
minimum=-20, maximum=20
|
17 |
+
))
|
18 |
+
|
19 |
+
inputs.append(gr.Slider(
|
20 |
+
label="Rotate Yaw", info='''Rotation yaw: Adjusts the left and right turn of the face''', value=0,
|
21 |
+
minimum=-20, maximum=20
|
22 |
+
))
|
23 |
+
|
24 |
+
inputs.append(gr.Slider(
|
25 |
+
label="Rotate Roll", info='''Rotation roll: Adjusts the tilt of the face to the left or right''', value=0,
|
26 |
+
minimum=-20, maximum=20
|
27 |
+
))
|
28 |
+
|
29 |
+
inputs.append(gr.Slider(
|
30 |
+
label="Blink", info='''Blink: Controls the degree of eye closure''', value=0,
|
31 |
+
minimum=-20, maximum=5
|
32 |
+
))
|
33 |
+
|
34 |
+
inputs.append(gr.Slider(
|
35 |
+
label="Eyebrow", info='''Eyebrow: Adjusts the height and shape of the eyebrows''', value=0,
|
36 |
+
minimum=-10, maximum=15
|
37 |
+
))
|
38 |
+
|
39 |
+
inputs.append(gr.Number(
|
40 |
+
label="Wink", info='''Wink: Controls the degree of one eye closing''', value=0
|
41 |
+
))
|
42 |
+
|
43 |
+
inputs.append(gr.Slider(
|
44 |
+
label="Pupil X", info='''Pupil X: Adjusts the horizontal position of the pupils''', value=0,
|
45 |
+
minimum=-15, maximum=15
|
46 |
+
))
|
47 |
+
|
48 |
+
inputs.append(gr.Slider(
|
49 |
+
label="Pupil Y", info='''Pupil Y: Adjusts the vertical position of the pupils''', value=0,
|
50 |
+
minimum=-15, maximum=15
|
51 |
+
))
|
52 |
+
|
53 |
+
inputs.append(gr.Slider(
|
54 |
+
label="Aaa", info='''AAA: Controls the mouth opening for 'aaa' sound''', value=0,
|
55 |
+
minimum=-30, maximum=120
|
56 |
+
))
|
57 |
+
|
58 |
+
inputs.append(gr.Slider(
|
59 |
+
label="Eee", info='''EEE: Controls the mouth shape for 'eee' sound''', value=0,
|
60 |
+
minimum=-20, maximum=15
|
61 |
+
))
|
62 |
+
|
63 |
+
inputs.append(gr.Slider(
|
64 |
+
label="Woo", info='''WOO: Controls the mouth shape for 'woo' sound''', value=0,
|
65 |
+
minimum=-20, maximum=15
|
66 |
+
))
|
67 |
+
|
68 |
+
inputs.append(gr.Slider(
|
69 |
+
label="Smile", info='''Smile: Adjusts the degree of smiling''', value=0,
|
70 |
+
minimum=-0.3, maximum=1.3
|
71 |
+
))
|
72 |
+
|
73 |
+
inputs.append(gr.Number(
|
74 |
+
label="Src Ratio", info='''Source ratio''', value=1
|
75 |
+
))
|
76 |
+
|
77 |
+
inputs.append(gr.Slider(
|
78 |
+
label="Sample Ratio", info='''Sample ratio''', value=1,
|
79 |
+
minimum=-0.2, maximum=1.2
|
80 |
+
))
|
81 |
+
|
82 |
+
inputs.append(gr.Slider(
|
83 |
+
label="Crop Factor", info='''Crop factor''', value=1.7,
|
84 |
+
minimum=1.5, maximum=2.5
|
85 |
+
))
|
86 |
+
|
87 |
+
inputs.append(gr.Dropdown(
|
88 |
+
choices=['webp', 'jpg', 'png'], label="output_format", info='''Format of the output images''', value="webp"
|
89 |
+
))
|
90 |
+
|
91 |
+
inputs.append(gr.Number(
|
92 |
+
label="Output Quality", info='''Quality of the output images, from 0 to 100. 100 is best quality, 0 is lowest quality.''', value=95
|
93 |
+
))
|
94 |
+
|
95 |
+
names = ['image', 'rotate_pitch', 'rotate_yaw', 'rotate_roll', 'blink', 'eyebrow', 'wink', 'pupil_x', 'pupil_y', 'aaa', 'eee', 'woo', 'smile', 'src_ratio', 'sample_ratio', 'crop_factor', 'output_format', 'output_quality']
|
96 |
+
|
97 |
+
outputs = []
|
98 |
+
outputs.append(gr.Image())
|
99 |
+
|
100 |
+
expected_outputs = len(outputs)
|
101 |
+
def predict(request: gr.Request, *args, progress=gr.Progress(track_tqdm=True)):
|
102 |
+
headers = {'Content-Type': 'application/json'}
|
103 |
+
|
104 |
+
payload = {"input": {}}
|
105 |
+
|
106 |
+
|
107 |
+
parsed_url = urlparse(str(request.url))
|
108 |
+
base_url = parsed_url.scheme + "://" + parsed_url.netloc
|
109 |
+
for i, key in enumerate(names):
|
110 |
+
value = args[i]
|
111 |
+
if value and (os.path.exists(str(value))):
|
112 |
+
value = f"{base_url}/file=" + value
|
113 |
+
if value is not None and value != "":
|
114 |
+
payload["input"][key] = value
|
115 |
+
|
116 |
+
response = requests.post("http://0.0.0.0:5000/predictions", headers=headers, json=payload)
|
117 |
+
|
118 |
+
|
119 |
+
if response.status_code == 201:
|
120 |
+
follow_up_url = response.json()["urls"]["get"]
|
121 |
+
response = requests.get(follow_up_url, headers=headers)
|
122 |
+
while response.json()["status"] != "succeeded":
|
123 |
+
if response.json()["status"] == "failed":
|
124 |
+
raise gr.Error("The submission failed!")
|
125 |
+
response = requests.get(follow_up_url, headers=headers)
|
126 |
+
time.sleep(1)
|
127 |
+
if response.status_code == 200:
|
128 |
+
json_response = response.json()
|
129 |
+
#If the output component is JSON return the entire output response
|
130 |
+
if(outputs[0].get_config()["name"] == "json"):
|
131 |
+
return json_response["output"]
|
132 |
+
predict_outputs = parse_outputs(json_response["output"])
|
133 |
+
processed_outputs = process_outputs(predict_outputs)
|
134 |
+
difference_outputs = expected_outputs - len(processed_outputs)
|
135 |
+
# If less outputs than expected, hide the extra ones
|
136 |
+
if difference_outputs > 0:
|
137 |
+
extra_outputs = [gr.update(visible=False)] * difference_outputs
|
138 |
+
processed_outputs.extend(extra_outputs)
|
139 |
+
# If more outputs than expected, cap the outputs to the expected number
|
140 |
+
elif difference_outputs < 0:
|
141 |
+
processed_outputs = processed_outputs[:difference_outputs]
|
142 |
+
|
143 |
+
return tuple(processed_outputs) if len(processed_outputs) > 1 else processed_outputs[0]
|
144 |
+
else:
|
145 |
+
if(response.status_code == 409):
|
146 |
+
raise gr.Error(f"Sorry, the Cog image is still processing. Try again in a bit.")
|
147 |
+
raise gr.Error(f"The submission failed! Error: {response.status_code}")
|
148 |
+
|
149 |
+
title = "Demo for expression-editor cog image by fofr"
|
150 |
+
model_description = "None"
|
151 |
+
|
152 |
+
app = gr.Interface(
|
153 |
+
fn=predict,
|
154 |
+
inputs=inputs,
|
155 |
+
outputs=outputs,
|
156 |
+
title=title,
|
157 |
+
description=model_description,
|
158 |
+
allow_flagging="never",
|
159 |
+
)
|
160 |
+
app.launch(share=True)
|
161 |
+
|
requirements.txt
ADDED
@@ -0,0 +1,2 @@
|
|
|
|
|
|
|
1 |
+
gradio==4.18.0
|
2 |
+
prance
|
run.sh
ADDED
@@ -0,0 +1,44 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
# Start the cog server in the background - Ensure correct path to cog
|
2 |
+
cd /src && python3 -m cog.server.http --threads=10 &
|
3 |
+
|
4 |
+
# Initialize counter for the first loop
|
5 |
+
counter1=0
|
6 |
+
|
7 |
+
# Continuous loop for reliably checking cog server's readiness on port 5000
|
8 |
+
while true; do
|
9 |
+
if nc -z localhost 5000; then
|
10 |
+
echo "Cog server is running on port 5000."
|
11 |
+
break # Exit the loop when the server is up
|
12 |
+
fi
|
13 |
+
echo "Waiting for cog server to start on port 5000..."
|
14 |
+
sleep 5
|
15 |
+
((counter1++))
|
16 |
+
if [ $counter1 -ge 250 ]; then
|
17 |
+
echo "Error: Cog server did not start on port 5000 after 250 attempts."
|
18 |
+
exit 1 # Exit the script with an error status
|
19 |
+
fi
|
20 |
+
done
|
21 |
+
|
22 |
+
# Initialize counter for the second loop
|
23 |
+
counter2=0
|
24 |
+
|
25 |
+
# New check: Waiting for the cog server to be fully ready
|
26 |
+
while true; do
|
27 |
+
response=$(curl -s http://localhost:5000/health-check) # Replace localhost:5000 with actual hostname and port if necessary
|
28 |
+
status=$(echo $response | jq -r '.status') # Parse status from JSON response
|
29 |
+
if [ "$status" = "READY" ]; then
|
30 |
+
echo "Cog server is fully ready."
|
31 |
+
break # Exit the loop when the server is fully ready
|
32 |
+
else
|
33 |
+
echo "Waiting for cog server (models loading) on port 5000..."
|
34 |
+
sleep 5
|
35 |
+
fi
|
36 |
+
((counter2++))
|
37 |
+
if [ $counter2 -ge 250 ]; then
|
38 |
+
echo "Error: Cog server did not become fully ready after 250 attempts."
|
39 |
+
exit 1 # Exit the script with an error status
|
40 |
+
fi
|
41 |
+
done
|
42 |
+
|
43 |
+
# Run the application - only when cog server is fully ready
|
44 |
+
cd $HOME/app && . $HOME/.venv/bin/activate && python app.py
|
utils/gradio_helpers.py
ADDED
@@ -0,0 +1,469 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import gradio as gr
|
2 |
+
from urllib.parse import urlparse
|
3 |
+
import requests
|
4 |
+
import time
|
5 |
+
from PIL import Image
|
6 |
+
import base64
|
7 |
+
import io
|
8 |
+
import uuid
|
9 |
+
import os
|
10 |
+
|
11 |
+
|
12 |
+
def extract_property_info(prop):
|
13 |
+
combined_prop = {}
|
14 |
+
merge_keywords = ["allOf", "anyOf", "oneOf"]
|
15 |
+
|
16 |
+
for keyword in merge_keywords:
|
17 |
+
if keyword in prop:
|
18 |
+
for subprop in prop[keyword]:
|
19 |
+
combined_prop.update(subprop)
|
20 |
+
del prop[keyword]
|
21 |
+
|
22 |
+
if not combined_prop:
|
23 |
+
combined_prop = prop.copy()
|
24 |
+
|
25 |
+
for key in ["description", "default"]:
|
26 |
+
if key in prop:
|
27 |
+
combined_prop[key] = prop[key]
|
28 |
+
|
29 |
+
return combined_prop
|
30 |
+
|
31 |
+
|
32 |
+
def detect_file_type(filename):
|
33 |
+
audio_extensions = [".mp3", ".wav", ".flac", ".aac", ".ogg", ".m4a"]
|
34 |
+
image_extensions = [
|
35 |
+
".jpg",
|
36 |
+
".jpeg",
|
37 |
+
".png",
|
38 |
+
".gif",
|
39 |
+
".bmp",
|
40 |
+
".tiff",
|
41 |
+
".svg",
|
42 |
+
".webp",
|
43 |
+
]
|
44 |
+
video_extensions = [
|
45 |
+
".mp4",
|
46 |
+
".mov",
|
47 |
+
".wmv",
|
48 |
+
".flv",
|
49 |
+
".avi",
|
50 |
+
".avchd",
|
51 |
+
".mkv",
|
52 |
+
".webm",
|
53 |
+
]
|
54 |
+
|
55 |
+
# Extract the file extension
|
56 |
+
if isinstance(filename, str):
|
57 |
+
extension = filename[filename.rfind(".") :].lower()
|
58 |
+
|
59 |
+
# Check the extension against each list
|
60 |
+
if extension in audio_extensions:
|
61 |
+
return "audio"
|
62 |
+
elif extension in image_extensions:
|
63 |
+
return "image"
|
64 |
+
elif extension in video_extensions:
|
65 |
+
return "video"
|
66 |
+
else:
|
67 |
+
return "string"
|
68 |
+
elif isinstance(filename, list):
|
69 |
+
return "list"
|
70 |
+
|
71 |
+
|
72 |
+
def build_gradio_inputs(ordered_input_schema, example_inputs=None):
|
73 |
+
inputs = []
|
74 |
+
input_field_strings = """inputs = []\n"""
|
75 |
+
names = []
|
76 |
+
for index, (name, prop) in enumerate(ordered_input_schema):
|
77 |
+
names.append(name)
|
78 |
+
prop = extract_property_info(prop)
|
79 |
+
if "enum" in prop:
|
80 |
+
input_field = gr.Dropdown(
|
81 |
+
choices=prop["enum"],
|
82 |
+
label=prop.get("title"),
|
83 |
+
info=prop.get("description"),
|
84 |
+
value=prop.get("default"),
|
85 |
+
)
|
86 |
+
input_field_string = f"""inputs.append(gr.Dropdown(
|
87 |
+
choices={prop["enum"]}, label="{prop.get("title")}", info={"'''"+prop.get("description")+"'''" if prop.get("description") else 'None'}, value="{prop.get("default")}"
|
88 |
+
))\n"""
|
89 |
+
elif prop["type"] == "integer":
|
90 |
+
if prop.get("minimum") and prop.get("maximum"):
|
91 |
+
input_field = gr.Slider(
|
92 |
+
label=prop.get("title"),
|
93 |
+
info=prop.get("description"),
|
94 |
+
value=prop.get("default"),
|
95 |
+
minimum=prop.get("minimum"),
|
96 |
+
maximum=prop.get("maximum"),
|
97 |
+
step=1,
|
98 |
+
)
|
99 |
+
input_field_string = f"""inputs.append(gr.Slider(
|
100 |
+
label="{prop.get("title")}", info={"'''"+prop.get("description")+"'''" if prop.get("description") else 'None'}, value={prop.get("default")},
|
101 |
+
minimum={prop.get("minimum")}, maximum={prop.get("maximum")}, step=1,
|
102 |
+
))\n"""
|
103 |
+
else:
|
104 |
+
input_field = gr.Number(
|
105 |
+
label=prop.get("title"),
|
106 |
+
info=prop.get("description"),
|
107 |
+
value=prop.get("default"),
|
108 |
+
)
|
109 |
+
input_field_string = f"""inputs.append(gr.Number(
|
110 |
+
label="{prop.get("title")}", info={"'''"+prop.get("description")+"'''" if prop.get("description") else 'None'}, value={prop.get("default")}
|
111 |
+
))\n"""
|
112 |
+
elif prop["type"] == "number":
|
113 |
+
if prop.get("minimum") and prop.get("maximum"):
|
114 |
+
input_field = gr.Slider(
|
115 |
+
label=prop.get("title"),
|
116 |
+
info=prop.get("description"),
|
117 |
+
value=prop.get("default"),
|
118 |
+
minimum=prop.get("minimum"),
|
119 |
+
maximum=prop.get("maximum"),
|
120 |
+
)
|
121 |
+
input_field_string = f"""inputs.append(gr.Slider(
|
122 |
+
label="{prop.get("title")}", info={"'''"+prop.get("description")+"'''" if prop.get("description") else 'None'}, value={prop.get("default")},
|
123 |
+
minimum={prop.get("minimum")}, maximum={prop.get("maximum")}
|
124 |
+
))\n"""
|
125 |
+
else:
|
126 |
+
input_field = gr.Number(
|
127 |
+
label=prop.get("title"),
|
128 |
+
info=prop.get("description"),
|
129 |
+
value=prop.get("default"),
|
130 |
+
)
|
131 |
+
input_field_string = f"""inputs.append(gr.Number(
|
132 |
+
label="{prop.get("title")}", info={"'''"+prop.get("description")+"'''" if prop.get("description") else 'None'}, value={prop.get("default")}
|
133 |
+
))\n"""
|
134 |
+
elif prop["type"] == "boolean":
|
135 |
+
input_field = gr.Checkbox(
|
136 |
+
label=prop.get("title"),
|
137 |
+
info=prop.get("description"),
|
138 |
+
value=prop.get("default"),
|
139 |
+
)
|
140 |
+
input_field_string = f"""inputs.append(gr.Checkbox(
|
141 |
+
label="{prop.get("title")}", info={"'''"+prop.get("description")+"'''" if prop.get("description") else 'None'}, value={prop.get("default")}
|
142 |
+
))\n"""
|
143 |
+
elif (
|
144 |
+
prop["type"] == "string" and prop.get("format") == "uri" and example_inputs
|
145 |
+
):
|
146 |
+
input_type_example = example_inputs.get(name, None)
|
147 |
+
if input_type_example:
|
148 |
+
input_type = detect_file_type(input_type_example)
|
149 |
+
else:
|
150 |
+
input_type = None
|
151 |
+
if input_type == "image":
|
152 |
+
input_field = gr.Image(label=prop.get("title"), type="filepath")
|
153 |
+
input_field_string = f"""inputs.append(gr.Image(
|
154 |
+
label="{prop.get("title")}", type="filepath"
|
155 |
+
))\n"""
|
156 |
+
elif input_type == "audio":
|
157 |
+
input_field = gr.Audio(label=prop.get("title"), type="filepath")
|
158 |
+
input_field_string = f"""inputs.append(gr.Audio(
|
159 |
+
label="{prop.get("title")}", type="filepath"
|
160 |
+
))\n"""
|
161 |
+
elif input_type == "video":
|
162 |
+
input_field = gr.Video(label=prop.get("title"))
|
163 |
+
input_field_string = f"""inputs.append(gr.Video(
|
164 |
+
label="{prop.get("title")}"
|
165 |
+
))\n"""
|
166 |
+
else:
|
167 |
+
input_field = gr.File(label=prop.get("title"))
|
168 |
+
input_field_string = f"""inputs.append(gr.File(
|
169 |
+
label="{prop.get("title")}"
|
170 |
+
))\n"""
|
171 |
+
else:
|
172 |
+
input_field = gr.Textbox(
|
173 |
+
label=prop.get("title"),
|
174 |
+
info=prop.get("description"),
|
175 |
+
)
|
176 |
+
input_field_string = f"""inputs.append(gr.Textbox(
|
177 |
+
label="{prop.get("title")}", info={"'''"+prop.get("description")+"'''" if prop.get("description") else 'None'}
|
178 |
+
))\n"""
|
179 |
+
inputs.append(input_field)
|
180 |
+
input_field_strings += f"{input_field_string}\n"
|
181 |
+
|
182 |
+
input_field_strings += f"names = {names}\n"
|
183 |
+
|
184 |
+
return inputs, input_field_strings, names
|
185 |
+
|
186 |
+
|
187 |
+
def build_gradio_outputs_replicate(output_types):
|
188 |
+
outputs = []
|
189 |
+
output_field_strings = """outputs = []\n"""
|
190 |
+
if output_types:
|
191 |
+
for output in output_types:
|
192 |
+
if output == "image":
|
193 |
+
output_field = gr.Image()
|
194 |
+
output_field_string = "outputs.append(gr.Image())"
|
195 |
+
elif output == "audio":
|
196 |
+
output_field = gr.Audio(type="filepath")
|
197 |
+
output_field_string = "outputs.append(gr.Audio(type='filepath'))"
|
198 |
+
elif output == "video":
|
199 |
+
output_field = gr.Video()
|
200 |
+
output_field_string = "outputs.append(gr.Video())"
|
201 |
+
elif output == "string":
|
202 |
+
output_field = gr.Textbox()
|
203 |
+
output_field_string = "outputs.append(gr.Textbox())"
|
204 |
+
elif output == "json":
|
205 |
+
output_field = gr.JSON()
|
206 |
+
output_field_string = "outputs.append(gr.JSON())"
|
207 |
+
elif output == "list":
|
208 |
+
output_field = gr.JSON()
|
209 |
+
output_field_string = "outputs.append(gr.JSON())"
|
210 |
+
outputs.append(output_field)
|
211 |
+
output_field_strings += f"{output_field_string}\n"
|
212 |
+
else:
|
213 |
+
output_field = gr.JSON()
|
214 |
+
output_field_string = "outputs.append(gr.JSON())"
|
215 |
+
outputs.append(output_field)
|
216 |
+
|
217 |
+
return outputs, output_field_strings
|
218 |
+
|
219 |
+
|
220 |
+
def build_gradio_outputs_cog():
|
221 |
+
pass
|
222 |
+
|
223 |
+
|
224 |
+
def process_outputs(outputs):
|
225 |
+
output_values = []
|
226 |
+
for output in outputs:
|
227 |
+
if not output:
|
228 |
+
continue
|
229 |
+
if isinstance(output, str):
|
230 |
+
if output.startswith("data:image"):
|
231 |
+
base64_data = output.split(",", 1)[1]
|
232 |
+
image_data = base64.b64decode(base64_data)
|
233 |
+
image_stream = io.BytesIO(image_data)
|
234 |
+
image = Image.open(image_stream)
|
235 |
+
output_values.append(image)
|
236 |
+
elif output.startswith("data:audio"):
|
237 |
+
base64_data = output.split(",", 1)[1]
|
238 |
+
audio_data = base64.b64decode(base64_data)
|
239 |
+
audio_stream = io.BytesIO(audio_data)
|
240 |
+
filename = f"{uuid.uuid4()}.wav" # Change format as needed
|
241 |
+
with open(filename, "wb") as audio_file:
|
242 |
+
audio_file.write(audio_stream.getbuffer())
|
243 |
+
output_values.append(filename)
|
244 |
+
elif output.startswith("data:video"):
|
245 |
+
base64_data = output.split(",", 1)[1]
|
246 |
+
video_data = base64.b64decode(base64_data)
|
247 |
+
video_stream = io.BytesIO(video_data)
|
248 |
+
# Here you can save the audio or return the stream for further processing
|
249 |
+
filename = f"{uuid.uuid4()}.mp4" # Change format as needed
|
250 |
+
with open(filename, "wb") as video_file:
|
251 |
+
video_file.write(video_stream.getbuffer())
|
252 |
+
output_values.append(filename)
|
253 |
+
else:
|
254 |
+
output_values.append(output)
|
255 |
+
else:
|
256 |
+
output_values.append(output)
|
257 |
+
return output_values
|
258 |
+
|
259 |
+
|
260 |
+
def parse_outputs(data):
|
261 |
+
if isinstance(data, dict):
|
262 |
+
# Handle case where data is an object
|
263 |
+
dict_values = []
|
264 |
+
for value in data.values():
|
265 |
+
extracted_values = parse_outputs(value)
|
266 |
+
# For dict, we append instead of extend to maintain list structure within objects
|
267 |
+
if isinstance(value, list):
|
268 |
+
dict_values += [extracted_values]
|
269 |
+
else:
|
270 |
+
dict_values += extracted_values
|
271 |
+
return dict_values
|
272 |
+
elif isinstance(data, list):
|
273 |
+
# Handle case where data is an array
|
274 |
+
list_values = []
|
275 |
+
for item in data:
|
276 |
+
# Here we extend to flatten the list since we're already in an array context
|
277 |
+
list_values += parse_outputs(item)
|
278 |
+
return list_values
|
279 |
+
else:
|
280 |
+
# Handle primitive data types directly
|
281 |
+
return [data]
|
282 |
+
|
283 |
+
|
284 |
+
def create_dynamic_gradio_app(
|
285 |
+
inputs,
|
286 |
+
outputs,
|
287 |
+
api_url,
|
288 |
+
api_id=None,
|
289 |
+
replicate_token=None,
|
290 |
+
title="",
|
291 |
+
model_description="",
|
292 |
+
names=[],
|
293 |
+
local_base=False,
|
294 |
+
hostname="0.0.0.0",
|
295 |
+
):
|
296 |
+
expected_outputs = len(outputs)
|
297 |
+
|
298 |
+
def predict(request: gr.Request, *args, progress=gr.Progress(track_tqdm=True)):
|
299 |
+
payload = {"input": {}}
|
300 |
+
if api_id:
|
301 |
+
payload["version"] = api_id
|
302 |
+
parsed_url = urlparse(str(request.url))
|
303 |
+
if local_base:
|
304 |
+
base_url = f"http://{hostname}:7860"
|
305 |
+
else:
|
306 |
+
base_url = parsed_url.scheme + "://" + parsed_url.netloc
|
307 |
+
for i, key in enumerate(names):
|
308 |
+
value = args[i]
|
309 |
+
if value and (os.path.exists(str(value))):
|
310 |
+
value = f"{base_url}/file=" + value
|
311 |
+
if value is not None and value != "":
|
312 |
+
payload["input"][key] = value
|
313 |
+
print(payload)
|
314 |
+
headers = {"Content-Type": "application/json"}
|
315 |
+
if replicate_token:
|
316 |
+
headers["Authorization"] = f"Token {replicate_token}"
|
317 |
+
print(headers)
|
318 |
+
response = requests.post(api_url, headers=headers, json=payload)
|
319 |
+
if response.status_code == 201:
|
320 |
+
follow_up_url = response.json()["urls"]["get"]
|
321 |
+
response = requests.get(follow_up_url, headers=headers)
|
322 |
+
while response.json()["status"] != "succeeded":
|
323 |
+
if response.json()["status"] == "failed":
|
324 |
+
raise gr.Error("The submission failed!")
|
325 |
+
response = requests.get(follow_up_url, headers=headers)
|
326 |
+
time.sleep(1)
|
327 |
+
# TODO: Add a failing mechanism if the API gets stuck
|
328 |
+
if response.status_code == 200:
|
329 |
+
json_response = response.json()
|
330 |
+
# If the output component is JSON return the entire output response
|
331 |
+
if outputs[0].get_config()["name"] == "json":
|
332 |
+
return json_response["output"]
|
333 |
+
predict_outputs = parse_outputs(json_response["output"])
|
334 |
+
processed_outputs = process_outputs(predict_outputs)
|
335 |
+
difference_outputs = expected_outputs - len(processed_outputs)
|
336 |
+
# If less outputs than expected, hide the extra ones
|
337 |
+
if difference_outputs > 0:
|
338 |
+
extra_outputs = [gr.update(visible=False)] * difference_outputs
|
339 |
+
processed_outputs.extend(extra_outputs)
|
340 |
+
# If more outputs than expected, cap the outputs to the expected number if
|
341 |
+
elif difference_outputs < 0:
|
342 |
+
processed_outputs = processed_outputs[:difference_outputs]
|
343 |
+
|
344 |
+
return (
|
345 |
+
tuple(processed_outputs)
|
346 |
+
if len(processed_outputs) > 1
|
347 |
+
else processed_outputs[0]
|
348 |
+
)
|
349 |
+
|
350 |
+
else:
|
351 |
+
if response.status_code == 409:
|
352 |
+
raise gr.Error(
|
353 |
+
f"Sorry, the Cog image is still processing. Try again in a bit."
|
354 |
+
)
|
355 |
+
raise gr.Error(f"The submission failed! Error: {response.status_code}")
|
356 |
+
|
357 |
+
app = gr.Interface(
|
358 |
+
fn=predict,
|
359 |
+
inputs=inputs,
|
360 |
+
outputs=outputs,
|
361 |
+
title=title,
|
362 |
+
description=model_description,
|
363 |
+
allow_flagging="never",
|
364 |
+
)
|
365 |
+
return app
|
366 |
+
|
367 |
+
|
368 |
+
def create_gradio_app_script(
|
369 |
+
inputs_string,
|
370 |
+
outputs_string,
|
371 |
+
api_url,
|
372 |
+
api_id=None,
|
373 |
+
replicate_token=None,
|
374 |
+
title="",
|
375 |
+
model_description="",
|
376 |
+
local_base=False,
|
377 |
+
hostname="0.0.0.0"
|
378 |
+
):
|
379 |
+
headers = {"Content-Type": "application/json"}
|
380 |
+
if replicate_token:
|
381 |
+
headers["Authorization"] = f"Token {replicate_token}"
|
382 |
+
|
383 |
+
if local_base:
|
384 |
+
base_url = f'base_url = "http://{hostname}:7860"'
|
385 |
+
else:
|
386 |
+
base_url = """parsed_url = urlparse(str(request.url))
|
387 |
+
base_url = parsed_url.scheme + "://" + parsed_url.netloc"""
|
388 |
+
headers_string = f"""headers = {headers}\n"""
|
389 |
+
api_id_value = f'payload["version"] = "{api_id}"' if api_id is not None else ""
|
390 |
+
definition_string = """expected_outputs = len(outputs)
|
391 |
+
def predict(request: gr.Request, *args, progress=gr.Progress(track_tqdm=True)):"""
|
392 |
+
payload_string = f"""payload = {{"input": {{}}}}
|
393 |
+
{api_id_value}
|
394 |
+
|
395 |
+
{base_url}
|
396 |
+
for i, key in enumerate(names):
|
397 |
+
value = args[i]
|
398 |
+
if value and (os.path.exists(str(value))):
|
399 |
+
value = f"{{base_url}}/file=" + value
|
400 |
+
if value is not None and value != "":
|
401 |
+
payload["input"][key] = value\n"""
|
402 |
+
|
403 |
+
request_string = (
|
404 |
+
f"""response = requests.post("{api_url}", headers=headers, json=payload)\n"""
|
405 |
+
)
|
406 |
+
|
407 |
+
result_string = f"""
|
408 |
+
if response.status_code == 201:
|
409 |
+
follow_up_url = response.json()["urls"]["get"]
|
410 |
+
response = requests.get(follow_up_url, headers=headers)
|
411 |
+
while response.json()["status"] != "succeeded":
|
412 |
+
if response.json()["status"] == "failed":
|
413 |
+
raise gr.Error("The submission failed!")
|
414 |
+
response = requests.get(follow_up_url, headers=headers)
|
415 |
+
time.sleep(1)
|
416 |
+
if response.status_code == 200:
|
417 |
+
json_response = response.json()
|
418 |
+
#If the output component is JSON return the entire output response
|
419 |
+
if(outputs[0].get_config()["name"] == "json"):
|
420 |
+
return json_response["output"]
|
421 |
+
predict_outputs = parse_outputs(json_response["output"])
|
422 |
+
processed_outputs = process_outputs(predict_outputs)
|
423 |
+
difference_outputs = expected_outputs - len(processed_outputs)
|
424 |
+
# If less outputs than expected, hide the extra ones
|
425 |
+
if difference_outputs > 0:
|
426 |
+
extra_outputs = [gr.update(visible=False)] * difference_outputs
|
427 |
+
processed_outputs.extend(extra_outputs)
|
428 |
+
# If more outputs than expected, cap the outputs to the expected number
|
429 |
+
elif difference_outputs < 0:
|
430 |
+
processed_outputs = processed_outputs[:difference_outputs]
|
431 |
+
|
432 |
+
return tuple(processed_outputs) if len(processed_outputs) > 1 else processed_outputs[0]
|
433 |
+
else:
|
434 |
+
if(response.status_code == 409):
|
435 |
+
raise gr.Error(f"Sorry, the Cog image is still processing. Try again in a bit.")
|
436 |
+
raise gr.Error(f"The submission failed! Error: {{response.status_code}}")\n"""
|
437 |
+
|
438 |
+
interface_string = f"""title = "{title}"
|
439 |
+
model_description = "{model_description}"
|
440 |
+
|
441 |
+
app = gr.Interface(
|
442 |
+
fn=predict,
|
443 |
+
inputs=inputs,
|
444 |
+
outputs=outputs,
|
445 |
+
title=title,
|
446 |
+
description=model_description,
|
447 |
+
allow_flagging="never",
|
448 |
+
)
|
449 |
+
app.launch(share=True)
|
450 |
+
"""
|
451 |
+
|
452 |
+
app_string = f"""import gradio as gr
|
453 |
+
from urllib.parse import urlparse
|
454 |
+
import requests
|
455 |
+
import time
|
456 |
+
import os
|
457 |
+
|
458 |
+
from utils.gradio_helpers import parse_outputs, process_outputs
|
459 |
+
|
460 |
+
{inputs_string}
|
461 |
+
{outputs_string}
|
462 |
+
{definition_string}
|
463 |
+
{headers_string}
|
464 |
+
{payload_string}
|
465 |
+
{request_string}
|
466 |
+
{result_string}
|
467 |
+
{interface_string}
|
468 |
+
"""
|
469 |
+
return app_string
|