You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
When I attempted to train the multi-decoder version with default config (final_RoP_Cov_A_fMCG.yaml), the traing process would be interrputed by infinity error at first epoch.
Setting trainable_cov to False will not help solve it.
If chaning running_mean_mode to "sliding", the training loss will be huge negative, and it becomes quite unstable
Did this happen when you train the network? And is this normal? Thank you in advance
The text was updated successfully, but these errors were encountered:
Hi @stepankonev,
When I attempted to train the multi-decoder version with default config (final_RoP_Cov_A_fMCG.yaml), the traing process would be interrputed by infinity error at first epoch.
Setting trainable_cov to False will not help solve it.
If chaning running_mean_mode to "sliding", the training loss will be huge negative, and it becomes quite unstable
Did this happen when you train the network? And is this normal? Thank you in advance
The text was updated successfully, but these errors were encountered: