/
NAFNet-width32.yml
108 lines (92 loc) · 2.13 KB
/
NAFNet-width32.yml
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
# ------------------------------------------------------------------------
# Copyright (c) 2022 megvii-model. All Rights Reserved.
# ------------------------------------------------------------------------
# Modified from BasicSR (https://github.com/xinntao/BasicSR)
# Copyright 2018-2020 BasicSR Authors
# ------------------------------------------------------------------------
# general settings
name: NAFNet-SIDD-width32
model_type: ImageRestorationModel
scale: 1
num_gpu: 8
manual_seed: 10
datasets:
train:
name: SIDD
type: PairedImageDataset
dataroot_gt: ./datasets/SIDD/train/gt_crops.lmdb
dataroot_lq: ./datasets/SIDD/train/input_crops.lmdb
filename_tmpl: '{}'
io_backend:
type: lmdb
gt_size: 256
use_flip: false
use_rot: false
# data loader
use_shuffle: true
num_worker_per_gpu: 4
batch_size_per_gpu: 4
dataset_enlarge_ratio: 1
prefetch_mode: ~
val:
name: SIDD_val
type: PairedImageDataset
dataroot_gt: ./datasets/SIDD/val/gt_crops.lmdb
dataroot_lq: ./datasets/SIDD/val/input_crops.lmdb
io_backend:
type: lmdb
network_g:
type: NAFNet
width: 32
enc_blk_nums: [2, 2, 4, 8]
middle_blk_num: 12
dec_blk_nums: [2, 2, 2, 2]
# path
path:
pretrain_network_g: ~
strict_load_g: true
resume_state: ~
# training settings
train:
optim_g:
type: AdamW
lr: !!float 1e-3
weight_decay: 0.
betas: [0.9, 0.9]
scheduler:
type: TrueCosineAnnealingLR
T_max: 200000
eta_min: !!float 1e-7
total_iter: 200000
warmup_iter: -1 # no warm up
# losses
pixel_opt:
type: PSNRLoss
loss_weight: 1
reduction: mean
# validation settings
val:
val_freq: !!float 2e4
save_img: false
use_image: false
metrics:
psnr: # metric name, can be arbitrary
type: calculate_psnr
crop_border: 0
test_y_channel: false
ssim:
type: calculate_ssim
crop_border: 0
test_y_channel: false
# logging settings
logger:
print_freq: 200
save_checkpoint_freq: !!float 5e3
use_tb_logger: true
wandb:
project: ~
resume_id: ~
# dist training settings
dist_params:
backend: nccl
port: 29500