Description

• Updated AutoEncoder to use eval_batch_size instead of batch_size (training batch size) for calculating loss stats in get_anomaly_score_losses. Using eval_batch_size is more appropriate since it will not impact the final trained model and affect model convergence. The default eval_batch_size (1024) is larger than default batch_size (256) and results in a speedup in training postproc ([BUG]: Long production DFP training times #816)
• Fixed a couple debug statements that were throwing string formatting errors.

Checklist

[x] I am familiar with the Contributing Guidelines.
[x] New or existing tests cover these changes.
[x] The documentation is up to date with these changes.