mmpretrain/configs/spark/spark_sparse-resnet50_8xb512-amp-coslr-1600e_in1k.py
Yixiao Fang a1cfe888e2
[Feature] Support SparK. (#1531)
* add spark configs

* fix configs

* remove repeat aug

* add module codes

* support lr layer decay of resnet

* update

* fix lint

* add metafile and readme

* fix lint

* add models and logs

* refactor codes

* fix lint

* update model rst

* update name

* add docstring

* add ut

* fix lint

---------

Co-authored-by: Ma Zerun <mzr1996@163.com>
2023-06-19 11:27:50 +08:00

31 lines
669 B
Python

_base_ = 'spark_sparse-resnet50_8xb512-amp-coslr-800e_in1k.py'
# learning rate scheduler
param_scheduler = [
dict(
type='LinearLR',
start_factor=1e-4,
by_epoch=True,
begin=0,
end=40,
convert_to_iter_based=True),
dict(
type='CosineAnnealingLR',
T_max=1560,
by_epoch=True,
begin=40,
end=1600,
convert_to_iter_based=True),
dict(
type='CosineAnnealingWeightDecay',
eta_min=0.2,
T_max=1600,
by_epoch=True,
begin=0,
end=1600,
convert_to_iter_based=True)
]
# runtime settings
train_cfg = dict(max_epochs=1600)