-
Notifications
You must be signed in to change notification settings - Fork 1.1k
/
riformer-s24_8xb128_in1k.py
39 lines (36 loc) · 1.06 KB
/
riformer-s24_8xb128_in1k.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
_base_ = [
'../_base_/datasets/imagenet_bs128_poolformer_small_224.py',
'../_base_/schedules/imagenet_bs1024_adamw_swin.py',
'../_base_/default_runtime.py',
]
# Model settings
model = dict(
type='ImageClassifier',
backbone=dict(
type='RIFormer',
arch='s24',
drop_path_rate=0.1,
init_cfg=[
dict(
type='TruncNormal',
layer=['Conv2d', 'Linear'],
std=.02,
bias=0.),
dict(type='Constant', layer=['GroupNorm'], val=1., bias=0.),
]),
neck=dict(type='GlobalAveragePooling'),
head=dict(
type='LinearClsHead',
num_classes=1000,
in_channels=512,
loss=dict(type='CrossEntropyLoss', loss_weight=1.0),
))
# schedule settings
optim_wrapper = dict(
optimizer=dict(lr=4e-3),
clip_grad=dict(max_norm=5.0),
)
# NOTE: `auto_scale_lr` is for automatically scaling LR
# based on the actual training batch size.
# base_batch_size = (32 GPUs) x (128 samples per GPU)
auto_scale_lr = dict(base_batch_size=4096)