fix: out of memory error
Browse files
app.py
CHANGED
@@ -10,6 +10,8 @@ if gr.NO_RELOAD:
|
|
10 |
high_noise_frac = 0.8
|
11 |
negative_prompt = "Distorted, discontinuous, Ugly, blurry, low resolution, motionless, static, disfigured, disconnected limbs, Ugly faces, incomplete arms"
|
12 |
generator = torch.manual_seed(8888)
|
|
|
|
|
13 |
|
14 |
base = DiffusionPipeline.from_pretrained(
|
15 |
"stabilityai/stable-diffusion-xl-base-1.0",
|
@@ -35,18 +37,21 @@ if gr.NO_RELOAD:
|
|
35 |
refiner.unet = torch.compile(refiner.unet, mode="reduce-overhead", fullgraph=True)
|
36 |
pipeline.unet = torch.compile(pipeline.unet, mode="reduce-overhead", fullgraph=True)
|
37 |
|
38 |
-
def generate(prompt: str):
|
|
|
39 |
image = base(
|
40 |
prompt=prompt,
|
41 |
num_inference_steps=n_steps,
|
42 |
denoising_end=high_noise_frac,
|
43 |
output_type="latent",
|
|
|
44 |
).images[0]
|
45 |
image = refiner(
|
46 |
prompt=prompt,
|
47 |
num_inference_steps=n_steps,
|
48 |
denoising_start=high_noise_frac,
|
49 |
image=image,
|
|
|
50 |
).images[0]
|
51 |
image = to_tensor(image)
|
52 |
frames: list[Image.Image] = pipeline(
|
@@ -56,6 +61,8 @@ def generate(prompt: str):
|
|
56 |
negative_prompt=negative_prompt,
|
57 |
guidance_scale=9.0,
|
58 |
generator=generator,
|
|
|
|
|
59 |
).frames[0]
|
60 |
frames = [to_tensor(frame.convert("RGB")).mul(255).byte().permute(1, 2, 0) for frame in frames]
|
61 |
frames = torch.stack(frames)
|
|
|
10 |
high_noise_frac = 0.8
|
11 |
negative_prompt = "Distorted, discontinuous, Ugly, blurry, low resolution, motionless, static, disfigured, disconnected limbs, Ugly faces, incomplete arms"
|
12 |
generator = torch.manual_seed(8888)
|
13 |
+
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
|
14 |
+
print("Device:", device)
|
15 |
|
16 |
base = DiffusionPipeline.from_pretrained(
|
17 |
"stabilityai/stable-diffusion-xl-base-1.0",
|
|
|
37 |
refiner.unet = torch.compile(refiner.unet, mode="reduce-overhead", fullgraph=True)
|
38 |
pipeline.unet = torch.compile(pipeline.unet, mode="reduce-overhead", fullgraph=True)
|
39 |
|
40 |
+
def generate(prompt: str, progress=gr.Progress()):
|
41 |
+
progress((0, 100), desc="Starting..")
|
42 |
image = base(
|
43 |
prompt=prompt,
|
44 |
num_inference_steps=n_steps,
|
45 |
denoising_end=high_noise_frac,
|
46 |
output_type="latent",
|
47 |
+
callback=lambda s, t, d: progress((s, 100), desc="Generating first frame..."),
|
48 |
).images[0]
|
49 |
image = refiner(
|
50 |
prompt=prompt,
|
51 |
num_inference_steps=n_steps,
|
52 |
denoising_start=high_noise_frac,
|
53 |
image=image,
|
54 |
+
callback=lambda s, t, d: progress((s+40, 100), desc="Refining first frame..."),
|
55 |
).images[0]
|
56 |
image = to_tensor(image)
|
57 |
frames: list[Image.Image] = pipeline(
|
|
|
61 |
negative_prompt=negative_prompt,
|
62 |
guidance_scale=9.0,
|
63 |
generator=generator,
|
64 |
+
decode_chunk_size=10,
|
65 |
+
callback=lambda s, t, d: progress((s+50, 100), desc="Generating video..."),
|
66 |
).frames[0]
|
67 |
frames = [to_tensor(frame.convert("RGB")).mul(255).byte().permute(1, 2, 0) for frame in frames]
|
68 |
frames = torch.stack(frames)
|