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
In both Megatron and Megatron LoRA my gradient norm during training is 0, all predictions are zeros. Maybe some problem with weights initialization or gradient accumulation?
The text was updated successfully, but these errors were encountered:
In both Megatron and Megatron LoRA my gradient norm during training is 0, all predictions are zeros. Maybe some problem with weights initialization or gradient accumulation?
The text was updated successfully, but these errors were encountered: