Open mkabra opened 2 years ago
If training fails for any reason, it is treated as trained for NaN iterations. Maybe have the ability to completely delete the failed training and revert to the previously trained model state or no tracker trained state.
If training fails for any reason, it is treated as trained for NaN iterations. Maybe have the ability to completely delete the failed training and revert to the previously trained model state or no tracker trained state.