forked from open-mmlab/mmpretrain
-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathdeit-base_16xb32_in1k-384px.py
37 lines (33 loc) · 999 Bytes
/
deit-base_16xb32_in1k-384px.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
_base_ = [
'../_base_/datasets/imagenet_bs64_swin_384.py',
'../_base_/schedules/imagenet_bs4096_AdamW.py',
'../_base_/default_runtime.py'
]
# model settings
model = dict(
type='ImageClassifier',
backbone=dict(
type='VisionTransformer',
arch='deit-base',
img_size=384,
patch_size=16,
),
neck=None,
head=dict(
type='VisionTransformerClsHead',
num_classes=1000,
in_channels=768,
loss=dict(
type='LabelSmoothLoss', label_smooth_val=0.1, mode='original'),
),
# Change to the path of the pretrained model
# init_cfg=dict(type='Pretrained', checkpoint=''),
)
# dataset settings
train_dataloader = dict(batch_size=32)
# schedule settings
optim_wrapper = dict(clip_grad=dict(max_norm=1.0))
# NOTE: `auto_scale_lr` is for automatically scaling LR
# based on the actual training batch size.
# base_batch_size = (16 GPUs) x (32 samples per GPU)
auto_scale_lr = dict(base_batch_size=512)