Spaces:
Runtime error
Runtime error
File size: 8,186 Bytes
b57c333 7e0c74d b57c333 acac4ca b57c333 7e0c74d b57c333 acac4ca b57c333 7e0c74d b57c333 7e0c74d b57c333 7e0c74d b57c333 acac4ca b57c333 7e0c74d b57c333 acac4ca b57c333 7e0c74d b57c333 acac4ca b57c333 acac4ca b57c333 6aa163b 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 |
#!/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)
with gr.Row():
with gr.Column():
with gr.Box():
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)
data_path = gr.Dropdown(
label='data path',
choices=[
'FateZero/data/teaser_car-turn',
'FateZero/data/style/sunflower',
# add shape editing ckpt here
],
value='FateZero/data/teaser_car-turn')
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')
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.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')
with gr.Row():
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,
# 50,
# 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,
],
outputs=result,
fn=merge_config_then_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,
]
# prompt.submit(fn=pipe.run, inputs=inputs, outputs=result)
target_prompt.submit(fn=merge_config_then_run, inputs=inputs, outputs=result)
# run_button.click(fn=pipe.run, inputs=inputs, outputs=result)
run_button.click(fn=merge_config_then_run, inputs=inputs, outputs=result)
demo.queue().launch()
|