Spaces:
Runtime error
Runtime error
Upload app.py
Browse files
app.py
ADDED
@@ -0,0 +1,373 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import torch
|
2 |
+
import gradio as gr
|
3 |
+
from gradio.themes.utils import colors, fonts, sizes
|
4 |
+
import argparse
|
5 |
+
from omegaconf import OmegaConf
|
6 |
+
import os
|
7 |
+
from models import get_models
|
8 |
+
from diffusers.utils.import_utils import is_xformers_available
|
9 |
+
from tca.tca_transform import tca_transform_model
|
10 |
+
from diffusers.models import AutoencoderKL
|
11 |
+
from models.clip import TextEmbedder
|
12 |
+
from datasets import video_transforms
|
13 |
+
from torchvision import transforms
|
14 |
+
from utils import mask_generation_before
|
15 |
+
from backend import auto_inpainting
|
16 |
+
from einops import rearrange
|
17 |
+
import torchvision
|
18 |
+
import sys
|
19 |
+
from PIL import Image
|
20 |
+
from ip_adapter.ip_adapter_transform import ip_scale_set, ip_transform_model
|
21 |
+
from transformers import CLIPVisionModelWithProjection, CLIPImageProcessor
|
22 |
+
from transformers.image_transforms import convert_to_rgb
|
23 |
+
try:
|
24 |
+
import utils
|
25 |
+
|
26 |
+
from diffusion import create_diffusion
|
27 |
+
from download import find_model
|
28 |
+
except:
|
29 |
+
# sys.path.append(os.getcwd())
|
30 |
+
sys.path.append(os.path.split(sys.path[0])[0])
|
31 |
+
# 代码解释
|
32 |
+
# sys.path[0] : 得到C:\Users\maxu\Desktop\blog_test\pakage2
|
33 |
+
# os.path.split(sys.path[0]) : 得到['C:\Users\maxu\Desktop\blog_test',pakage2']
|
34 |
+
# mmcls 里面跨包引用是因为安装了mmcls
|
35 |
+
|
36 |
+
import utils
|
37 |
+
|
38 |
+
from diffusion import create_diffusion
|
39 |
+
from download import find_model
|
40 |
+
|
41 |
+
|
42 |
+
def auto_inpainting(video_input, masked_video, mask, prompt, image, vae, text_encoder, image_encoder, diffusion, model, device, cfg_scale, img_cfg_scale, negative_prompt=""):
|
43 |
+
global use_fp16
|
44 |
+
image_prompt_embeds = None
|
45 |
+
if prompt is None:
|
46 |
+
prompt = ""
|
47 |
+
if image is not None:
|
48 |
+
clip_image = clip_image_processor(images=image, return_tensors="pt").pixel_values
|
49 |
+
clip_image_embeds = image_encoder(clip_image.to(device)).image_embeds
|
50 |
+
uncond_clip_image_embeds = torch.zeros_like(clip_image_embeds).to(device)
|
51 |
+
image_prompt_embeds = torch.cat([clip_image_embeds, uncond_clip_image_embeds], dim=0)
|
52 |
+
image_prompt_embeds = rearrange(image_prompt_embeds, '(b n) c -> b n c', b=2).contiguous()
|
53 |
+
model = ip_scale_set(model, img_cfg_scale)
|
54 |
+
if use_fp16:
|
55 |
+
image_prompt_embeds = image_prompt_embeds.to(dtype=torch.float16)
|
56 |
+
b, f, c, h, w = video_input.shape
|
57 |
+
latent_h = video_input.shape[-2] // 8
|
58 |
+
latent_w = video_input.shape[-1] // 8
|
59 |
+
|
60 |
+
if use_fp16:
|
61 |
+
z = torch.randn(1, 4, 16, latent_h, latent_w, dtype=torch.float16, device=device) # b,c,f,h,w
|
62 |
+
masked_video = masked_video.to(dtype=torch.float16)
|
63 |
+
mask = mask.to(dtype=torch.float16)
|
64 |
+
else:
|
65 |
+
z = torch.randn(1, 4, 16, latent_h, latent_w, device=device) # b,c,f,h,w
|
66 |
+
|
67 |
+
masked_video = rearrange(masked_video, 'b f c h w -> (b f) c h w').contiguous()
|
68 |
+
masked_video = vae.encode(masked_video).latent_dist.sample().mul_(0.18215)
|
69 |
+
masked_video = rearrange(masked_video, '(b f) c h w -> b c f h w', b=b).contiguous()
|
70 |
+
mask = torch.nn.functional.interpolate(mask[:,:,0,:], size=(latent_h, latent_w)).unsqueeze(1)
|
71 |
+
masked_video = torch.cat([masked_video] * 2)
|
72 |
+
mask = torch.cat([mask] * 2)
|
73 |
+
z = torch.cat([z] * 2)
|
74 |
+
prompt_all = [prompt] + [negative_prompt]
|
75 |
+
|
76 |
+
text_prompt = text_encoder(text_prompts=prompt_all, train=False)
|
77 |
+
model_kwargs = dict(encoder_hidden_states=text_prompt,
|
78 |
+
class_labels=None,
|
79 |
+
cfg_scale=cfg_scale,
|
80 |
+
use_fp16=use_fp16,
|
81 |
+
ip_hidden_states=image_prompt_embeds)
|
82 |
+
|
83 |
+
# Sample images:
|
84 |
+
samples = diffusion.ddim_sample_loop(
|
85 |
+
model.forward_with_cfg, z.shape, z, clip_denoised=False, model_kwargs=model_kwargs, progress=True, device=device, \
|
86 |
+
mask=mask, x_start=masked_video, use_concat=True
|
87 |
+
)
|
88 |
+
samples, _ = samples.chunk(2, dim=0) # [1, 4, 16, 32, 32]
|
89 |
+
if use_fp16:
|
90 |
+
samples = samples.to(dtype=torch.float16)
|
91 |
+
|
92 |
+
video_clip = samples[0].permute(1, 0, 2, 3).contiguous() # [16, 4, 32, 32]
|
93 |
+
video_clip = vae.decode(video_clip / 0.18215).sample # [16, 3, 256, 256]
|
94 |
+
return video_clip
|
95 |
+
|
96 |
+
|
97 |
+
def auto_inpainting_temp_split(video_input, masked_video, mask, prompt, image, vae, text_encoder, image_encoder, diffusion, model, device, scfg_scale, tcfg_scale, img_cfg_scale, negative_prompt=""):
|
98 |
+
global use_fp16
|
99 |
+
image_prompt_embeds = None
|
100 |
+
if prompt is None:
|
101 |
+
prompt = ""
|
102 |
+
if image is not None:
|
103 |
+
clip_image = clip_image_processor(images=image, return_tensors="pt").pixel_values
|
104 |
+
clip_image_embeds = image_encoder(clip_image.to(device)).image_embeds
|
105 |
+
uncond_clip_image_embeds = torch.zeros_like(clip_image_embeds).to(device)
|
106 |
+
image_prompt_embeds = torch.cat([clip_image_embeds, clip_image_embeds, uncond_clip_image_embeds], dim=0)
|
107 |
+
image_prompt_embeds = rearrange(image_prompt_embeds, '(b n) c -> b n c', b=3).contiguous()
|
108 |
+
model = ip_scale_set(model, img_cfg_scale)
|
109 |
+
if use_fp16:
|
110 |
+
image_prompt_embeds = image_prompt_embeds.to(dtype=torch.float16)
|
111 |
+
b, f, c, h, w = video_input.shape
|
112 |
+
latent_h = video_input.shape[-2] // 8
|
113 |
+
latent_w = video_input.shape[-1] // 8
|
114 |
+
|
115 |
+
if use_fp16:
|
116 |
+
z = torch.randn(1, 4, 16, latent_h, latent_w, dtype=torch.float16, device=device) # b,c,f,h,w
|
117 |
+
masked_video = masked_video.to(dtype=torch.float16)
|
118 |
+
mask = mask.to(dtype=torch.float16)
|
119 |
+
else:
|
120 |
+
z = torch.randn(1, 4, 16, latent_h, latent_w, device=device) # b,c,f,h,w
|
121 |
+
|
122 |
+
masked_video = rearrange(masked_video, 'b f c h w -> (b f) c h w').contiguous()
|
123 |
+
masked_video = vae.encode(masked_video).latent_dist.sample().mul_(0.18215)
|
124 |
+
masked_video = rearrange(masked_video, '(b f) c h w -> b c f h w', b=b).contiguous()
|
125 |
+
mask = torch.nn.functional.interpolate(mask[:,:,0,:], size=(latent_h, latent_w)).unsqueeze(1)
|
126 |
+
masked_video = torch.cat([masked_video] * 3)
|
127 |
+
mask = torch.cat([mask] * 3)
|
128 |
+
z = torch.cat([z] * 3)
|
129 |
+
prompt_all = [prompt] + [prompt] + [negative_prompt]
|
130 |
+
prompt_temp = [prompt] + [""] + [""]
|
131 |
+
|
132 |
+
text_prompt = text_encoder(text_prompts=prompt_all, train=False)
|
133 |
+
temporal_text_prompt = text_encoder(text_prompts=prompt_temp, train=False)
|
134 |
+
model_kwargs = dict(encoder_hidden_states=text_prompt,
|
135 |
+
class_labels=None,
|
136 |
+
scfg_scale=scfg_scale,
|
137 |
+
tcfg_scale=tcfg_scale,
|
138 |
+
use_fp16=use_fp16,
|
139 |
+
ip_hidden_states=image_prompt_embeds,
|
140 |
+
encoder_temporal_hidden_states=temporal_text_prompt)
|
141 |
+
|
142 |
+
# Sample images:
|
143 |
+
samples = diffusion.ddim_sample_loop(
|
144 |
+
model.forward_with_cfg_temp_split, z.shape, z, clip_denoised=False, model_kwargs=model_kwargs, progress=True, device=device, \
|
145 |
+
mask=mask, x_start=masked_video, use_concat=True
|
146 |
+
)
|
147 |
+
samples, _ = samples.chunk(2, dim=0) # [1, 4, 16, 32, 32]
|
148 |
+
if use_fp16:
|
149 |
+
samples = samples.to(dtype=torch.float16)
|
150 |
+
|
151 |
+
video_clip = samples[0].permute(1, 0, 2, 3).contiguous() # [16, 4, 32, 32]
|
152 |
+
video_clip = vae.decode(video_clip / 0.18215).sample # [16, 3, 256, 256]
|
153 |
+
return video_clip
|
154 |
+
|
155 |
+
|
156 |
+
# ========================================
|
157 |
+
# Model Initialization
|
158 |
+
# ========================================
|
159 |
+
device = None
|
160 |
+
output_path = None
|
161 |
+
use_fp16 = False
|
162 |
+
model = None
|
163 |
+
vae = None
|
164 |
+
text_encoder = None
|
165 |
+
image_encoder = None
|
166 |
+
clip_image_processor = None
|
167 |
+
def init_model():
|
168 |
+
global device
|
169 |
+
global output_path
|
170 |
+
global use_fp16
|
171 |
+
global model
|
172 |
+
global diffusion
|
173 |
+
global vae
|
174 |
+
global text_encoder
|
175 |
+
global image_encoder
|
176 |
+
global clip_image_processor
|
177 |
+
print('Initializing ShowMaker', flush=True)
|
178 |
+
parser = argparse.ArgumentParser()
|
179 |
+
parser.add_argument("--config", type=str, default="./configs/sample_mask.yaml")
|
180 |
+
args = parser.parse_args()
|
181 |
+
args = OmegaConf.load(args.config)
|
182 |
+
device = "cuda" if torch.cuda.is_available() else "cpu"
|
183 |
+
output_path = args.save_img_path
|
184 |
+
# Load model:
|
185 |
+
latent_h = args.image_size[0] // 8
|
186 |
+
latent_w = args.image_size[1] // 8
|
187 |
+
args.image_h = args.image_size[0]
|
188 |
+
args.image_w = args.image_size[1]
|
189 |
+
args.latent_h = latent_h
|
190 |
+
args.latent_w = latent_w
|
191 |
+
print('loading model')
|
192 |
+
model = get_models(True, args).to(device)
|
193 |
+
model = tca_transform_model(model).to(device)
|
194 |
+
model = ip_transform_model(model).to(device)
|
195 |
+
if args.use_compile:
|
196 |
+
model = torch.compile(model)
|
197 |
+
if args.enable_xformers_memory_efficient_attention:
|
198 |
+
if is_xformers_available():
|
199 |
+
model.enable_xformers_memory_efficient_attention()
|
200 |
+
print("xformer!")
|
201 |
+
else:
|
202 |
+
raise ValueError("xformers is not available. Make sure it is installed correctly")
|
203 |
+
ckpt_path = args.ckpt
|
204 |
+
state_dict = find_model(ckpt_path)
|
205 |
+
model.load_state_dict(state_dict)
|
206 |
+
print('loading succeed')
|
207 |
+
model.eval() # important!
|
208 |
+
pretrained_model_path = args.pretrained_model_path
|
209 |
+
vae = AutoencoderKL.from_pretrained(pretrained_model_path, subfolder="vae").to(device)
|
210 |
+
text_encoder = TextEmbedder(tokenizer_path=pretrained_model_path + "tokenizer",
|
211 |
+
encoder_path=pretrained_model_path + "text_encoder").to(device)
|
212 |
+
image_encoder = CLIPVisionModelWithProjection.from_pretrained(args.image_encoder_path).to(device)
|
213 |
+
clip_image_processor = CLIPImageProcessor()
|
214 |
+
if args.use_fp16:
|
215 |
+
print('Warnning: using half percision for inferencing!')
|
216 |
+
vae.to(dtype=torch.float16)
|
217 |
+
model.to(dtype=torch.float16)
|
218 |
+
text_encoder.to(dtype=torch.float16)
|
219 |
+
image_encoder.to(dtype=torch.float16)
|
220 |
+
use_fp16 = True
|
221 |
+
print('Initialization Finished')
|
222 |
+
init_model()
|
223 |
+
|
224 |
+
|
225 |
+
# ========================================
|
226 |
+
# Video Generation
|
227 |
+
# ========================================
|
228 |
+
def video_generation(text, image, scfg_scale, tcfg_scale, img_cfg_scale, diffusion):
|
229 |
+
with torch.no_grad():
|
230 |
+
print("begin generation", flush=True)
|
231 |
+
transform_video = transforms.Compose([
|
232 |
+
video_transforms.ToTensorVideo(), # TCHW
|
233 |
+
video_transforms.WebVideo320512((320, 512)),
|
234 |
+
transforms.Normalize(mean=[0.5, 0.5, 0.5], std=[0.5, 0.5, 0.5], inplace=True)
|
235 |
+
])
|
236 |
+
video_frames = torch.zeros(16, 3, 320, 512, dtype=torch.uint8)
|
237 |
+
video_frames = transform_video(video_frames)
|
238 |
+
video_input = video_frames.to(device).unsqueeze(0) # b,f,c,h,w
|
239 |
+
mask = mask_generation_before("all", video_input.shape, video_input.dtype, device)
|
240 |
+
masked_video = video_input * (mask == 0)
|
241 |
+
if image is not None:
|
242 |
+
print(image.shape, flush=True)
|
243 |
+
# image = Image.open(image)
|
244 |
+
if scfg_scale == tcfg_scale:
|
245 |
+
video_clip = auto_inpainting(video_input, masked_video, mask, text, image, vae, text_encoder, image_encoder, diffusion, model, device, scfg_scale, img_cfg_scale)
|
246 |
+
else:
|
247 |
+
video_clip = auto_inpainting_temp_split(video_input, masked_video, mask, text, image, vae, text_encoder, image_encoder, diffusion, model, device, scfg_scale, tcfg_scale, img_cfg_scale)
|
248 |
+
video_clip = ((video_clip * 0.5 + 0.5) * 255).add_(0.5).clamp_(0, 255).to(dtype=torch.uint8).cpu().permute(0, 2, 3, 1)
|
249 |
+
video_path = os.path.join(output_path, 'video.mp4')
|
250 |
+
torchvision.io.write_video(video_path, video_clip, fps=8)
|
251 |
+
return video_path
|
252 |
+
|
253 |
+
|
254 |
+
# ========================================
|
255 |
+
# Video Prediction
|
256 |
+
# ========================================
|
257 |
+
def video_prediction(text, image, scfg_scale, tcfg_scale, img_cfg_scale, preframe, diffusion):
|
258 |
+
with torch.no_grad():
|
259 |
+
print("begin generation", flush=True)
|
260 |
+
transform_video = transforms.Compose([
|
261 |
+
video_transforms.ToTensorVideo(), # TCHW
|
262 |
+
# video_transforms.WebVideo320512((320, 512)),
|
263 |
+
transforms.Normalize(mean=[0.5, 0.5, 0.5], std=[0.5, 0.5, 0.5], inplace=True)
|
264 |
+
])
|
265 |
+
preframe = torch.as_tensor(convert_to_rgb(preframe)).unsqueeze(0)
|
266 |
+
zeros = torch.zeros_like(preframe)
|
267 |
+
video_frames = torch.cat([preframe] + [zeros] * 15, dim=0).permute(0, 3, 1, 2)
|
268 |
+
H_scale = 320 / video_frames.shape[2]
|
269 |
+
W_scale = 512 / video_frames.shape[3]
|
270 |
+
scale_ = H_scale
|
271 |
+
if W_scale < H_scale:
|
272 |
+
scale_ = W_scale
|
273 |
+
video_frames = torch.nn.functional.interpolate(video_frames, scale_factor=scale_, mode="bilinear", align_corners=False)
|
274 |
+
video_frames = transform_video(video_frames)
|
275 |
+
video_input = video_frames.to(device).unsqueeze(0) # b,f,c,h,w
|
276 |
+
mask = mask_generation_before("first1", video_input.shape, video_input.dtype, device)
|
277 |
+
masked_video = video_input * (mask == 0)
|
278 |
+
if image is not None:
|
279 |
+
print(image.shape, flush=True)
|
280 |
+
if scfg_scale == tcfg_scale:
|
281 |
+
video_clip = auto_inpainting(video_input, masked_video, mask, text, image, vae, text_encoder, image_encoder, diffusion, model, device, scfg_scale, img_cfg_scale)
|
282 |
+
else:
|
283 |
+
video_clip = auto_inpainting_temp_split(video_input, masked_video, mask, text, image, vae, text_encoder, image_encoder, diffusion, model, device, scfg_scale, tcfg_scale, img_cfg_scale)
|
284 |
+
video_clip = ((video_clip * 0.5 + 0.5) * 255).add_(0.5).clamp_(0, 255).to(dtype=torch.uint8).cpu().permute(0, 2, 3, 1)
|
285 |
+
video_path = os.path.join(output_path, 'video.mp4')
|
286 |
+
torchvision.io.write_video(video_path, video_clip, fps=8)
|
287 |
+
return video_path
|
288 |
+
|
289 |
+
|
290 |
+
# ========================================
|
291 |
+
# Judge Generation or Prediction
|
292 |
+
# ========================================
|
293 |
+
def gen_or_pre(text_input, image_input, scfg_scale, tcfg_scale, img_cfg_scale, preframe_input, diffusion_step):
|
294 |
+
default_step = [25, 40, 50, 100, 125, 200, 250]
|
295 |
+
difference = [abs(item - diffusion_step) for item in default_step]
|
296 |
+
diffusion_step = default_step[difference.index(min(difference))]
|
297 |
+
diffusion = create_diffusion(str(diffusion_step))
|
298 |
+
if preframe_input is None:
|
299 |
+
return video_generation(text_input, image_input, scfg_scale, tcfg_scale, img_cfg_scale, diffusion)
|
300 |
+
else:
|
301 |
+
return video_prediction(text_input, image_input, scfg_scale, tcfg_scale, img_cfg_scale, preframe_input, diffusion)
|
302 |
+
|
303 |
+
|
304 |
+
with gr.Blocks() as demo:
|
305 |
+
with gr.Row():
|
306 |
+
with gr.Column(visible=True) as input_raws:
|
307 |
+
with gr.Row():
|
308 |
+
with gr.Column(scale=1.0):
|
309 |
+
text_input = gr.Textbox(show_label=True, interactive=True, label="Text prompt").style(container=False)
|
310 |
+
with gr.Row():
|
311 |
+
with gr.Column(scale=0.5):
|
312 |
+
image_input = gr.Image(show_label=True, interactive=True, label="Reference image").style(container=False)
|
313 |
+
with gr.Column(scale=0.5):
|
314 |
+
preframe_input = gr.Image(show_label=True, interactive=True, label="First frame").style(container=False)
|
315 |
+
with gr.Row():
|
316 |
+
with gr.Column(scale=1.0):
|
317 |
+
scfg_scale = gr.Slider(
|
318 |
+
minimum=1,
|
319 |
+
maximum=50,
|
320 |
+
value=8,
|
321 |
+
step=0.1,
|
322 |
+
interactive=True,
|
323 |
+
label="Spatial Text Guidence Scale",
|
324 |
+
)
|
325 |
+
with gr.Row():
|
326 |
+
with gr.Column(scale=1.0):
|
327 |
+
tcfg_scale = gr.Slider(
|
328 |
+
minimum=1,
|
329 |
+
maximum=50,
|
330 |
+
value=6.5,
|
331 |
+
step=0.1,
|
332 |
+
interactive=True,
|
333 |
+
label="Temporal Text Guidence Scale",
|
334 |
+
)
|
335 |
+
with gr.Row():
|
336 |
+
with gr.Column(scale=1.0):
|
337 |
+
img_cfg_scale = gr.Slider(
|
338 |
+
minimum=0,
|
339 |
+
maximum=1,
|
340 |
+
value=0.3,
|
341 |
+
step=0.005,
|
342 |
+
interactive=True,
|
343 |
+
label="Image Guidence Scale",
|
344 |
+
)
|
345 |
+
with gr.Row():
|
346 |
+
with gr.Column(scale=1.0):
|
347 |
+
diffusion_step = gr.Slider(
|
348 |
+
minimum=20,
|
349 |
+
maximum=250,
|
350 |
+
value=100,
|
351 |
+
step=1,
|
352 |
+
interactive=True,
|
353 |
+
label="Diffusion Step",
|
354 |
+
)
|
355 |
+
with gr.Row():
|
356 |
+
with gr.Column(scale=0.5, min_width=0):
|
357 |
+
run = gr.Button("💭Send")
|
358 |
+
with gr.Column(scale=0.5, min_width=0):
|
359 |
+
clear = gr.Button("🔄Clear️")
|
360 |
+
with gr.Column(scale=0.5, visible=True) as video_upload:
|
361 |
+
output_video = gr.Video(interactive=False, include_audio=True, elem_id="输出的视频")#.style(height=360)
|
362 |
+
# with gr.Column(elem_id="image", scale=0.5) as img_part:
|
363 |
+
# with gr.Tab("Video", elem_id='video_tab'):
|
364 |
+
|
365 |
+
# with gr.Tab("Image", elem_id='image_tab'):
|
366 |
+
# up_image = gr.Image(type="pil", interactive=True, elem_id="image_upload").style(height=360)
|
367 |
+
# upload_button = gr.Button(value="Upload & Start Chat", interactive=True, variant="primary")
|
368 |
+
clear = gr.Button("Restart")
|
369 |
+
run.click(gen_or_pre, [text_input, image_input, scfg_scale, tcfg_scale, img_cfg_scale, preframe_input, diffusion_step], [output_video])
|
370 |
+
|
371 |
+
# demo.launch(share=True, enable_queue=True)
|
372 |
+
|
373 |
+
demo.launch(server_name="0.0.0.0", server_port=10034, enable_queue=True)
|