Spaces:
Runtime error
Runtime error
File size: 13,958 Bytes
b57c333 0c86f09 b57c333 4dff355 b57c333 0c86f09 b57c333 afd7574 b57c333 afd7574 b57c333 afd7574 b57c333 acac4ca b57c333 7e0c74d b57c333 acac4ca b57c333 acac4ca b57c333 17a220f b57c333 afd7574 b57c333 17a220f b57c333 0c86f09 b57c333 4dff355 b57c333 afd7574 b57c333 4dff355 b57c333 4dff355 b57c333 |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 |
#!/usr/bin/env python
from __future__ import annotations
import os
import gradio as gr
# from inference import InferencePipeline
# from FateZero import test_fatezero
from inference_fatezero import merge_config_then_run
# class InferenceUtil:
# def __init__(self, hf_token: str | None):
# self.hf_token = hf_token
# def load_model_info(self, model_id: str) -> tuple[str, str]:
# # todo FIXME
# try:
# card = InferencePipeline.get_model_card(model_id, self.hf_token)
# except Exception:
# return '', ''
# base_model = getattr(card.data, 'base_model', '')
# training_prompt = getattr(card.data, 'training_prompt', '')
# return base_model, training_prompt
# TITLE = '# [FateZero](http://fate-zero-edit.github.io/)'
HF_TOKEN = os.getenv('HF_TOKEN')
# pipe = InferencePipeline(HF_TOKEN)
pipe = merge_config_then_run()
# app = InferenceUtil(HF_TOKEN)
with gr.Blocks(css='style.css') as demo:
# gr.Markdown(TITLE)
gr.HTML(
"""
<div style="text-align: center; max-width: 1200px; margin: 20px auto;">
<h1 style="font-weight: 900; font-size: 3rem; margin: 0rem">
FateZero : Fusing Attentions for Zero-shot Text-based Video Editing
</h1>
<h2 style="font-weight: 450; font-size: 1rem; margin: 0rem">
<a href="https://chenyangqiqi.github.io/">Chenyang Qi</a>
<a href="https://vinthony.github.io/academic/">Xiaodong Cun</a> , <a href="https://yzhang2016.github.io/">Yong Zhang</a>,
<a href="https://chenyanglei.github.io">Chenyang Lei</a>, <a href="https://xinntao.github.io/"> Xintao Wang </a>,
<a href="https://scholar.google.com/citations?user=4oXBp9UAAAAJ&hl=zh-CN">Ying Shan</a>,
<a href="http://cqf.io">Qifeng Chen</a>
</h2>
<h2 style="font-weight: 450; font-size: 1rem; margin: 0rem">
<span class="link-block">
[<a href="https://arxiv.org/abs/2303.09535" target="_blank"
class="external-link ">
<span class="icon">
<i class="ai ai-arxiv"></i>
</span>
<span>arXiv</span>
</a>]
</span>
<!-- Github link -->
<span class="link-block">
[<a href="https://github.com/ChenyangQiQi/FateZero" target="_blank"
class="external-link ">
<span class="icon">
<i class="fab fa-github"></i>
</span>
<span>Code</span>
</a>]
</span>
<!-- Github link -->
<span class="link-block">
[<a href="http://fate-zero-edit.github.io/" target="_blank"
class="external-link ">
<span class="icon">
<i class="fab fa-github"></i>
</span>
<span>Homepage</span>
</a>]
</span>
<!-- Github link -->
<span class="link-block">
[<a href="https://hkustconnect-my.sharepoint.com/:v:/g/personal/cqiaa_connect_ust_hk/EXKDI_nahEhKtiYPvvyU9SkBDTG2W4G1AZ_vkC7ekh3ENw?e=ficp9t" target="_blank"
class="external-link ">
<span class="icon">
<i class="fab fa-youtube"></i>
</span>
<span>Video</span>
</a>]
</span>
</h2>
<h2 style="font-weight: 450; font-size: 1rem; margin-top: 0.5rem; margin-bottom: 0.5rem">
FateZero is a first zero-shot framework for text-driven video editing via pretrained diffusion models without training.
</h2>
</div>
""")
gr.HTML("""
<p>For faster inference without waiting in queue, you may duplicate the space and upgrade to GPU in settings.
<br/>
<a href="https://huggingface.co/spaces/chenyangqi/FateZero?duplicate=true">
<img style="margin-top: 0em; margin-bottom: 0em" src="https://bit.ly/3gLdBN6" alt="Duplicate Space"></a>
</p>""")
with gr.Row():
with gr.Column():
with gr.Accordion('Input Video', open=True):
user_input_video = gr.File(label='Input Source Video')
with gr.Accordion('Temporal Crop offset and Sampling Stride', open=False):
n_sample_frame = gr.Slider(label='Number of Frames in Video',
minimum=0,
maximum=32,
step=1,
value=8)
stride = gr.Slider(label='Temporal sampling stride in Video',
minimum=0,
maximum=20,
step=1,
value=1)
start_sample_frame = gr.Number(label='Start frame in the video',
value=0,
precision=0)
with gr.Accordion('Spatial Crop offset', open=False):
left_crop = gr.Number(label='Left crop',
value=0,
precision=0)
right_crop = gr.Number(label='Right crop',
value=0,
precision=0)
top_crop = gr.Number(label='Top crop',
value=0,
precision=0)
bottom_crop = gr.Number(label='Bottom crop',
value=0,
precision=0)
offset_list = [
left_crop,
right_crop,
top_crop,
bottom_crop,
]
ImageSequenceDataset_list = [
start_sample_frame,
n_sample_frame,
stride
] + offset_list
data_path = gr.Dropdown(
label='provided data path',
choices=[
'FateZero/data/teaser_car-turn',
'FateZero/data/style/sunflower',
# add shape editing ckpt here
],
value='FateZero/data/teaser_car-turn')
model_id = gr.Dropdown(
label='Model ID',
choices=[
'CompVis/stable-diffusion-v1-4',
# add shape editing ckpt here
],
value='CompVis/stable-diffusion-v1-4')
# with gr.Accordion(
# label=
# 'Model info (Base model and prompt used for training)',
# open=False):
# with gr.Row():
# base_model_used_for_training = gr.Text(
# label='Base model', interactive=False)
# prompt_used_for_training = gr.Text(
# label='Training prompt', interactive=False)
with gr.Accordion('Text Prompt', open=True):
source_prompt = gr.Textbox(label='Source Prompt',
info='A good prompt describes each frame and most objects in video. Especially, it has the object or attribute that we want to edit or preserve.',
max_lines=1,
placeholder='Example: "a silver jeep driving down a curvy road in the countryside"',
value='a silver jeep driving down a curvy road in the countryside')
target_prompt = gr.Textbox(label='Target Prompt',
info='A reasonable composition of video may achieve better results(e.g., "sunflower" video with "Van Gogh" prompt is better than "sunflower" with "Monet")',
max_lines=1,
placeholder='Example: "watercolor painting of a silver jeep driving down a curvy road in the countryside"',
value='watercolor painting of a silver jeep driving down a curvy road in the countryside')
with gr.Accordion('DDIM Parameters', open=True):
num_steps = gr.Slider(label='Number of Steps',
info='larger value has better editing capacity, but takes more time and memory',
minimum=0,
maximum=50,
step=1,
value=10)
guidance_scale = gr.Slider(label='CFG Scale',
minimum=0,
maximum=50,
step=0.1,
value=7.5)
run_button = gr.Button('Generate')
# gr.Markdown('''
# - It takes a few minutes to download model first.
# - Expected time to generate an 8-frame video: 70 seconds with T4, 24 seconds with A10G, (10 seconds with A100)
# ''')
# gr.Markdown('''
# todo
# ''')
with gr.Column():
result = gr.Video(label='Result')
result.style(height=512, width=512)
with gr.Accordion('FateZero Parameters for attention fusing', open=True):
cross_replace_steps = gr.Slider(label='cross-attention replace steps',
info='More steps, replace more cross attention to preserve semantic layout.',
minimum=0.0,
maximum=1.0,
step=0.1,
value=0.7)
self_replace_steps = gr.Slider(label='self-attention replace steps',
info='More steps, replace more spatial-temporal self-attention to preserve geometry and motion.',
minimum=0.0,
maximum=1.0,
step=0.1,
value=0.7)
enhance_words = gr.Textbox(label='words to be enhanced',
info='Amplify the target-words cross attention',
max_lines=1,
placeholder='Example: "watercolor "',
value='watercolor')
enhance_words_value = gr.Slider(label='Amplify the target cross-attention',
info='larger value, more elements of target words',
minimum=0.0,
maximum=20.0,
step=1,
value=10)
with gr.Row():
from example import style_example
examples = style_example
# examples = [
# [
# 'CompVis/stable-diffusion-v1-4',
# 'FateZero/data/teaser_car-turn',
# 'a silver jeep driving down a curvy road in the countryside',
# 'watercolor painting of a silver jeep driving down a curvy road in the countryside',
# 0.8,
# 0.8,
# "watercolor",
# 10,
# 10,
# 7.5,
# ],
# [
# 'CompVis/stable-diffusion-v1-4',
# 'FateZero/data/style/sunflower',
# 'a yellow sunflower',
# 'van gogh style painting of a yellow sunflower',
# 0.5,
# 0.5,
# 'van gogh',
# 10,
# 10,
# 7.5,
# ],
# ]
gr.Examples(examples=examples,
inputs=[
model_id,
data_path,
source_prompt,
target_prompt,
cross_replace_steps,
self_replace_steps,
enhance_words,
enhance_words_value,
num_steps,
guidance_scale,
user_input_video,
*ImageSequenceDataset_list
],
outputs=result,
fn=pipe.run,
cache_examples=os.getenv('SYSTEM') == 'spaces')
# model_id.change(fn=app.load_model_info,
# inputs=model_id,
# outputs=[
# base_model_used_for_training,
# prompt_used_for_training,
# ])
inputs = [
model_id,
data_path,
source_prompt,
target_prompt,
cross_replace_steps,
self_replace_steps,
enhance_words,
enhance_words_value,
num_steps,
guidance_scale,
user_input_video,
*ImageSequenceDataset_list
]
# prompt.submit(fn=pipe.run, inputs=inputs, outputs=result)
target_prompt.submit(fn=pipe.run, inputs=inputs, outputs=result)
# run_button.click(fn=pipe.run, inputs=inputs, outputs=result)
run_button.click(fn=pipe.run, inputs=inputs, outputs=result)
demo.queue().launch()
|