What is the difference between "LearningRateDecayOptimizerConstructor" with the default one?

[Shiweiliuiiiiiii]

Dear authors,

Do I have to use ''LearningRateDecayOptimizerConstructor'' to reproduce the results of ConvNeXt on segmentation?

Can I use the default one but with the same decay parameters like this?

optimizer = dict(delete=True, type='AdamW',
lr=0.0001, betas=(0.9, 0.999), weight_decay=0.05,
paramwise_cfg={'decay_rate': 0.9,
'decay_type': 'stage_wise',
'num_layers': 12})

Many thanks,
Shiwei

[yxchng]

any updates?