detect what's causing NaNs

thushv89 / AdaCNN

AdaCNN algorithm. Clean implementation

0 stars 0 forks source link

detect what's causing NaNs #25

Closed thushv89 closed 6 years ago

thushv89 commented 6 years ago

Possible causes: L2 decay: No Adaptive dropout: No Pool Momentum:No Way params removed: Data distribution

Currently, Set use_dropout to False use_l2_loss to False pool_momentum = 0.0 rm_indices are a continuous block no finetune after remove action This seems to be the cause of NaNs

thushv89 commented 6 years ago

Possibly several causes; "slice_optimize", "learning rate"