_base_ = ['./regnetx-400mf_8xb128_in1k.py'] # model settings model = dict( backbone=dict(type='RegNet', arch='regnetx_12gf'), head=dict(in_channels=2240, )) # for batch_size 512, use lr = 0.4 optimizer = dict(lr=0.4) data = dict(samples_per_gpu=64, )
