-
Notifications
You must be signed in to change notification settings - Fork 0
/
eval.py
60 lines (47 loc) · 1.93 KB
/
eval.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
import os
from absl import app
import torch
from torchvision.utils import save_image, make_grid
from model import UNet
from config import FLAGS
from helpers import evaluate
from ddim import GaussianDiffusionTimestepsSampler
from hyper_diffusion import MomentumSampler
def load_checkpoint(path):
checkpoint = torch.load(path)
return {k.replace('.module', ''): v for k, v in checkpoint.items()}
def eval(argv):
model = UNet(
T=FLAGS.T, ch=FLAGS.ch, ch_mult=FLAGS.ch_mult, attn=FLAGS.attn,
num_res_blocks=FLAGS.num_res_blocks, dropout=FLAGS.dropout)
if FLAGS.model_checkpoint:
ckpt = torch.load(FLAGS.model_checkpoint)
ckpt = ckpt['net_model'] if 'net_model' in ckpt else ckpt
model.load_state_dict(ckpt)
model = model.cuda()
model.eval()
if FLAGS.sampler_type == 'ddim':
sampler = GaussianDiffusionTimestepsSampler(model,
FLAGS.beta_1,
FLAGS.beta_T,
T_orig = FLAGS.T,
T_reduced = FLAGS.T_reduced,
img_size=FLAGS.img_size,
mean_type=FLAGS.mean_type,
var_type=FLAGS.var_type).cuda()
sample_fn = lambda x: sampler.ddim_sample(x)
if FLAGS.sampler_type == 'momentum':
sampler = MomentumSampler(FLAGS.optimizer_time_steps).cuda()
if FLAGS.sampler_checkpoint:
print('Loading sampler checkpoint...')
checkpoint = load_checkpoint(FLAGS.sampler_checkpoint)
sampler.load_state_dict(checkpoint)
if FLAGS.time_embedding_checkpoint:
print('Loading checkpoint...')
checkpoint = load_checkpoint(FLAGS.time_embedding_checkpoint)
model.time_embedding.load_state_dict(checkpoint)
sample_fn = lambda x: sampler(model, x)
sampler.eval()
print(evaluate(sample_fn, save_images = True))
if __name__ == '__main__':
app.run(eval)