Doubts about lr .

[WEIZHIHONG720]

Hi, I see that the ircsn_ig65m_pretrained_bnfrozen_r152_32x2x1_58e_kinetics400_rgb.py file has an lr of 0.000125 and i would like to know if this value is the learning rate for fine tuning or for normal training? Thank you !

optimizer

optimizer = dict( type='SGD', lr=0.000125, momentum=0.9, weight_decay=0.0001) # this lr is used for 8 gpus

[Dai-Wenxun]

Hi, weizh588, The backbone is pretrained, so this learning rate should be used for fine-tuning.