-
Notifications
You must be signed in to change notification settings - Fork 0
/
denoising.py
149 lines (122 loc) · 4.99 KB
/
denoising.py
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
import hydra
from omegaconf import DictConfig, OmegaConf
from pathlib import Path
from tqdm import tqdm
import numpy as np
import torch
from src.trainer import GeneratorModule
from src.utils import get_config
def torch_to_image_numpy(tensor: torch.Tensor):
tensor = tensor * 0.5 + 0.5
im_np = [tensor[i].cpu().numpy().transpose(1, 2, 0) for i in range(tensor.shape[0])]
return im_np
@hydra.main(config_path="configs", config_name="config", version_base="1.3")
def main(cfg: DictConfig) -> None:
print(OmegaConf.to_yaml(cfg)) # Optional: To print the full config for debugging
device = "cuda" if torch.cuda.is_available() else "cpu"
runs_per_sample = cfg.generator.runs_per_sample
path_checkpoint = Path(cfg.trainer.checkpoint + cfg.generator.ckpt)
output_path = Path(
cfg.generator.output
+ f"\\{cfg.project_name}"
+ f"\\power{cfg['dataset']['power']}_{cfg['model']['loss']}_{cfg['generator']['generator_mode']}"
)
print(output_path)
output_path.mkdir(exist_ok=True, parents=True)
module = GeneratorModule.load_from_checkpoint(
checkpoint_path=path_checkpoint,
strict=False,
config=cfg,
use_fp16=cfg.trainer.fp16,
timestep_respacing=str(cfg.generator.timestep_respacing),
)
module = module.to(device)
module.eval()
diffusion = module.diffusion
model = module.model
model = model.to(device)
def model_fn(x_t, ts, **kwargs):
half = x_t[: len(x_t) // 2]
combined = torch.cat([half, half], dim=0)
model_out = model(combined, ts, **kwargs)
eps, rest = model_out[:, :3], model_out[:, 3:]
cond_eps, uncond_eps = torch.split(eps, len(eps) // 2, dim=0)
half_eps = uncond_eps + cfg.generator.guidance_scale * (cond_eps - uncond_eps)
eps = torch.cat([half_eps, half_eps], dim=0)
return torch.cat([eps, rest], dim=1)
shape = (
cfg.trainer.batch_size * 2,
module.n_channels,
module.size_image,
module.size_image,
)
generator_model = cfg["generator"]["generator_mode"]
# generate images
dls = [module.test_dataloader()]
names = ["test"]
if cfg.generator.include_val:
dls.append(module.val_dataloader())
names.append("val")
if cfg.generator.include_train:
dls.append(module.train_dataloader())
names.append("train")
try:
number_of_examples = (len(dls[0]) + len(dls[1])) * cfg.trainer.batch_size
except:
number_of_examples = len(dls[0]) * cfg.trainer.batch_size
print(f"Total number of examples: {number_of_examples}")
for i, dl in enumerate(dls):
for batch_idx, batch in tqdm(enumerate(dl), colour="blue"):
if batch_idx == cfg["generator"]["max_batch"]:
break
images = []
generated_images = []
dirty_noisy_list = []
sky_indexes_list = []
im_in = batch["true"]
im_in_ = im_in.to(device)
dirty_noisy = batch["dirty_noisy"].to(device)
filenames = batch["filename"]
for _ in tqdm(range(runs_per_sample)):
im_in = im_in_
with torch.no_grad():
zero_label_noise = torch.zeros_like(dirty_noisy, device=device)
dirty_noisy = torch.cat([dirty_noisy, zero_label_noise], dim=0)
if generator_model == "ddpm":
im_out = diffusion.p_sample_loop(
model_fn,
cond=dirty_noisy,
shape=shape,
device=device,
clip_denoised=True,
progress=False,
cond_fn=None,
)[: cfg.trainer.batch_size]
dirty_noisy = dirty_noisy[: cfg.trainer.batch_size]
else:
pass
im_in = torch_to_image_numpy(im_in)
im_out = torch_to_image_numpy(im_out)
dirty_noisy_ = torch_to_image_numpy(dirty_noisy)
images.extend(im_in)
generated_images.extend(im_out)
dirty_noisy_list.extend(dirty_noisy_)
sky_indexes_list.extend(filenames)
images = np.array(images)
generated_images = np.array(generated_images)
dirty_noisy_list = np.array(dirty_noisy_list)
np.save(output_path / f"batch={batch_idx}_{names[i]}_images.npy", images)
np.save(
output_path / f"batch={batch_idx}_{names[i]}_generated_images.npy",
generated_images,
)
np.save(
output_path / f"batch={batch_idx}_{names[i]}_dirty_noisy.npy",
dirty_noisy_list,
)
np.save(
output_path / f"batch={batch_idx}_{names[i]}_sky_indexes.npy",
sky_indexes_list,
)
if __name__ == "__main__":
main()