simsiam_resnet50_8xb32-coslr-200e_in1k.py 1.37 KB
Newer Older
renzhc's avatar
renzhc committed
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
_base_ = [
    '../_base_/datasets/imagenet_bs32_mocov2.py',
    '../_base_/schedules/imagenet_sgd_coslr_200e.py',
    '../_base_/default_runtime.py',
]

# model settings
model = dict(
    type='SimSiam',
    backbone=dict(
        type='ResNet',
        depth=50,
        norm_cfg=dict(type='SyncBN'),
        zero_init_residual=True),
    neck=dict(
        type='NonLinearNeck',
        in_channels=2048,
        hid_channels=2048,
        out_channels=2048,
        num_layers=3,
        with_last_bn_affine=False,
        with_avg_pool=True),
    head=dict(
        type='LatentPredictHead',
        loss=dict(type='CosineSimilarityLoss'),
        predictor=dict(
            type='NonLinearNeck',
            in_channels=2048,
            hid_channels=512,
            out_channels=2048,
            with_avg_pool=False,
            with_last_bn=False,
            with_last_bias=True)),
)

# optimizer
# set base learning rate
lr = 0.05
optim_wrapper = dict(
    type='OptimWrapper',
    optimizer=dict(type='SGD', lr=lr, weight_decay=1e-4, momentum=0.9),
    paramwise_cfg=dict(custom_keys={'predictor': dict(fix_lr=True)}))

# runtime settings
default_hooks = dict(
    # only keeps the latest 3 checkpoints
    checkpoint=dict(type='CheckpointHook', interval=10, max_keep_ckpts=3))

# additional hooks
custom_hooks = [
    dict(type='SimSiamHook', priority='HIGH', fix_pred_lr=True, lr=lr)
]