# optimizer optimizer = dict( type='SGD', lr=0.01, # this lr is used for 8 gpus momentum=0.9, weight_decay=0.0001) optimizer_config = dict(grad_clip=dict(max_norm=40, norm_type=2)) # learning policy lr_config = dict(policy='step', step=[40, 80]) total_epochs = 100
