Closed muyuuuu closed 2 years ago
what's meaning about 15, and why fill zero to parameters' gradient?
https://github.com/sahagobinda/GPM/blob/1a238ec9d2ca30bae8fd1707f161cc6bd093c72a/main_cifar100.py#L150-L157
K<15 is used to exclude gradient projections of the final classifier layer. fill_(0) is used to stop gradient updates of the batch-norm parameters in the model after task 1.
thank you~
what's meaning about 15, and why fill zero to parameters' gradient?
https://github.com/sahagobinda/GPM/blob/1a238ec9d2ca30bae8fd1707f161cc6bd093c72a/main_cifar100.py#L150-L157