Commit
·
1dade9e
1
Parent(s):
03c3c6a
Update base/app.py
Browse files- base/app.py +54 -22
base/app.py
CHANGED
|
@@ -15,12 +15,11 @@ args = OmegaConf.load("./base/configs/sample.yaml")
|
|
| 15 |
device = "cuda" if torch.cuda.is_available() else "cpu"
|
| 16 |
|
| 17 |
# ------- get model ---------------
|
| 18 |
-
model_t2V = model_t2v_fun(args)
|
| 19 |
-
model_t2V.to(device)
|
| 20 |
-
if device == "cuda":
|
| 21 |
-
|
| 22 |
|
| 23 |
-
# model_t2V.enable_xformers_memory_efficient_attention()
|
| 24 |
css = """
|
| 25 |
h1 {
|
| 26 |
text-align: center;
|
|
@@ -31,13 +30,46 @@ h1 {
|
|
| 31 |
}
|
| 32 |
"""
|
| 33 |
|
| 34 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 35 |
if seed_inp!=-1:
|
| 36 |
setup_seed(seed_inp)
|
| 37 |
else:
|
| 38 |
seed_inp = random.choice(range(10000000))
|
| 39 |
setup_seed(seed_inp)
|
| 40 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 41 |
print(videos[0].shape)
|
| 42 |
if not os.path.exists(args.output_folder):
|
| 43 |
os.mkdir(args.output_folder)
|
|
@@ -82,7 +114,7 @@ with gr.Blocks(css='style.css') as demo:
|
|
| 82 |
with gr.Column():
|
| 83 |
|
| 84 |
prompt = gr.Textbox(value="a corgi walking in the park at sunrise, oil painting style", label="Prompt", placeholder="enter prompt", show_label=True, elem_id="prompt-in", min_width=200, lines=2)
|
| 85 |
-
|
| 86 |
ddim_steps = gr.Slider(label='Steps', minimum=50, maximum=300, value=50, step=1)
|
| 87 |
seed_inp = gr.Slider(value=-1,label="seed (for random generation, use -1)",show_label=True,minimum=-1,maximum=2147483647)
|
| 88 |
cfg = gr.Number(label="guidance_scale",value=7.5)
|
|
@@ -94,24 +126,24 @@ with gr.Blocks(css='style.css') as demo:
|
|
| 94 |
clean_btn = gr.Button("Clean video")
|
| 95 |
video_out = gr.Video(label="Video result", elem_id="video-output")
|
| 96 |
|
| 97 |
-
inputs = [prompt, seed_inp, ddim_steps,cfg]
|
| 98 |
outputs = [video_out]
|
| 99 |
|
| 100 |
ex = gr.Examples(
|
| 101 |
-
examples = [['a corgi walking in the park at sunrise, oil painting style',400,50,7],
|
| 102 |
-
['a cut teddy bear reading a book in the park, oil painting style, high quality',700,50,7],
|
| 103 |
-
['an epic tornado attacking above a glowing city at night, the tornado is made of smoke, highly detailed',230,50,7],
|
| 104 |
-
['a jar filled with fire, 4K video, 3D rendered, well-rendered',400,50,7],
|
| 105 |
-
['a teddy bear walking in the park, oil painting style, high quality',400,50,7],
|
| 106 |
-
['a teddy bear walking on the street, 2k, high quality',100,50,7],
|
| 107 |
-
['a panda taking a selfie, 2k, high quality',400,50,7],
|
| 108 |
-
['a polar bear playing drum kit in NYC Times Square, 4k, high resolution',400,50,7],
|
| 109 |
-
['jungle river at sunset, ultra quality',400,50,7],
|
| 110 |
-
['a shark swimming in clear Carribean ocean, 2k, high quality',400,50,7],
|
| 111 |
-
['A steam train moving on a mountainside by Vincent van Gogh',230,50,7],
|
| 112 |
-
['a confused grizzly bear in calculus class',1000,50,7]],
|
| 113 |
fn = infer,
|
| 114 |
-
inputs=[prompt, seed_inp, ddim_steps,cfg],
|
| 115 |
outputs=[video_out],
|
| 116 |
cache_examples=False,
|
| 117 |
)
|
|
|
|
| 15 |
device = "cuda" if torch.cuda.is_available() else "cpu"
|
| 16 |
|
| 17 |
# ------- get model ---------------
|
| 18 |
+
# model_t2V = model_t2v_fun(args)
|
| 19 |
+
# model_t2V.to(device)
|
| 20 |
+
# if device == "cuda":
|
| 21 |
+
# model_t2V.enable_xformers_memory_efficient_attention()
|
| 22 |
|
|
|
|
| 23 |
css = """
|
| 24 |
h1 {
|
| 25 |
text-align: center;
|
|
|
|
| 30 |
}
|
| 31 |
"""
|
| 32 |
|
| 33 |
+
sd_path = args.pretrained_path + "/stable-diffusion-v1-4"
|
| 34 |
+
unet = get_models(args, sd_path).to(device, dtype=torch.float16)
|
| 35 |
+
state_dict = find_model("./pretrained_models/lavie_base.pt")
|
| 36 |
+
unet.load_state_dict(state_dict)
|
| 37 |
+
vae = AutoencoderKL.from_pretrained(sd_path, subfolder="vae", torch_dtype=torch.float16).to(device)
|
| 38 |
+
tokenizer_one = CLIPTokenizer.from_pretrained(sd_path, subfolder="tokenizer")
|
| 39 |
+
text_encoder_one = CLIPTextModel.from_pretrained(sd_path, subfolder="text_encoder", torch_dtype=torch.float16).to(device) # huge
|
| 40 |
+
unet.eval()
|
| 41 |
+
vae.eval()
|
| 42 |
+
text_encoder_one.eval()
|
| 43 |
+
|
| 44 |
+
def infer(prompt, seed_inp, ddim_steps,cfg, infer_type):
|
| 45 |
if seed_inp!=-1:
|
| 46 |
setup_seed(seed_inp)
|
| 47 |
else:
|
| 48 |
seed_inp = random.choice(range(10000000))
|
| 49 |
setup_seed(seed_inp)
|
| 50 |
+
if infer_type == 'ddim':
|
| 51 |
+
scheduler = DDIMScheduler.from_pretrained(sd_path,
|
| 52 |
+
subfolder="scheduler",
|
| 53 |
+
beta_start=args.beta_start,
|
| 54 |
+
beta_end=args.beta_end,
|
| 55 |
+
beta_schedule=args.beta_schedule)
|
| 56 |
+
elif infer_type == 'eulerdiscrete':
|
| 57 |
+
scheduler = EulerDiscreteScheduler.from_pretrained(sd_path,
|
| 58 |
+
subfolder="scheduler",
|
| 59 |
+
beta_start=args.beta_start,
|
| 60 |
+
beta_end=args.beta_end,
|
| 61 |
+
beta_schedule=args.beta_schedule)
|
| 62 |
+
elif infer_type == 'ddpm':
|
| 63 |
+
scheduler = DDPMScheduler.from_pretrained(sd_path,
|
| 64 |
+
subfolder="scheduler",
|
| 65 |
+
beta_start=args.beta_start,
|
| 66 |
+
beta_end=args.beta_end,
|
| 67 |
+
beta_schedule=args.beta_schedule)
|
| 68 |
+
model = VideoGenPipeline(vae=vae, text_encoder=text_encoder_one, tokenizer=tokenizer_one, scheduler=scheduler, unet=unet)
|
| 69 |
+
model.to(device)
|
| 70 |
+
if device == "cuda":
|
| 71 |
+
model.enable_xformers_memory_efficient_attention()
|
| 72 |
+
videos = model(prompt, video_length=16, height = 320, width= 512, num_inference_steps=ddim_steps, guidance_scale=cfg).video
|
| 73 |
print(videos[0].shape)
|
| 74 |
if not os.path.exists(args.output_folder):
|
| 75 |
os.mkdir(args.output_folder)
|
|
|
|
| 114 |
with gr.Column():
|
| 115 |
|
| 116 |
prompt = gr.Textbox(value="a corgi walking in the park at sunrise, oil painting style", label="Prompt", placeholder="enter prompt", show_label=True, elem_id="prompt-in", min_width=200, lines=2)
|
| 117 |
+
infer_type = gr.Dropdown(['ddpm','ddim','eulerdiscrete'], label='infer_type',value='ddim')
|
| 118 |
ddim_steps = gr.Slider(label='Steps', minimum=50, maximum=300, value=50, step=1)
|
| 119 |
seed_inp = gr.Slider(value=-1,label="seed (for random generation, use -1)",show_label=True,minimum=-1,maximum=2147483647)
|
| 120 |
cfg = gr.Number(label="guidance_scale",value=7.5)
|
|
|
|
| 126 |
clean_btn = gr.Button("Clean video")
|
| 127 |
video_out = gr.Video(label="Video result", elem_id="video-output")
|
| 128 |
|
| 129 |
+
inputs = [prompt, seed_inp, ddim_steps, cfg, infer_type]
|
| 130 |
outputs = [video_out]
|
| 131 |
|
| 132 |
ex = gr.Examples(
|
| 133 |
+
examples = [['a corgi walking in the park at sunrise, oil painting style',400,50,7,'ddim'],
|
| 134 |
+
['a cut teddy bear reading a book in the park, oil painting style, high quality',700,50,7,'ddim'],
|
| 135 |
+
['an epic tornado attacking above a glowing city at night, the tornado is made of smoke, highly detailed',230,50,7,'ddim'],
|
| 136 |
+
['a jar filled with fire, 4K video, 3D rendered, well-rendered',400,50,7,'ddim'],
|
| 137 |
+
['a teddy bear walking in the park, oil painting style, high quality',400,50,7,'ddim'],
|
| 138 |
+
['a teddy bear walking on the street, 2k, high quality',100,50,7,'ddim'],
|
| 139 |
+
['a panda taking a selfie, 2k, high quality',400,50,7,'ddim'],
|
| 140 |
+
['a polar bear playing drum kit in NYC Times Square, 4k, high resolution',400,50,7,'ddim'],
|
| 141 |
+
['jungle river at sunset, ultra quality',400,50,7,'ddim'],
|
| 142 |
+
['a shark swimming in clear Carribean ocean, 2k, high quality',400,50,7,'ddim'],
|
| 143 |
+
['A steam train moving on a mountainside by Vincent van Gogh',230,50,7,'ddim'],
|
| 144 |
+
['a confused grizzly bear in calculus class',1000,50,7,'ddim']],
|
| 145 |
fn = infer,
|
| 146 |
+
inputs=[prompt, seed_inp, ddim_steps,cfg,infer_type],
|
| 147 |
outputs=[video_out],
|
| 148 |
cache_examples=False,
|
| 149 |
)
|