backbone
gitdawn123 opened this issue · comments
dawn commented
hello,
you select ResNet-34 as the backbone of the Simulator. why is it?
is there any requirement for the backbone?
Chen Ma commented
ResNet-34 is not a very large network, so it can work well in meta-learning when using only a few data to fine-tune it. Besides, the fine-tuning requires time, and a small network will be faster.
dawn commented
i got it.
thanks!