This PR suggest to simple changes:
1) Allow the user to specify logging levels and an optional output path for log files in one line via utils.logger_settings()
2) Make the default output less versatile. In particular, we print metrics on the INFO level now once every epoch, instead of every 20 batches. The previous behaviour led to huge logs if the datasets had many (10k+) rows.
3) Compute training entropy after every epoch, not after every batch.
By submitting this pull request, I confirm that my contribution is made under the terms of the Apache 2.0 license.
This PR suggest to simple changes: 1) Allow the user to specify logging levels and an optional output path for log files in one line via
utils.logger_settings()2) Make the default output less versatile. In particular, we print metrics on the INFO level now once every epoch, instead of every 20 batches. The previous behaviour led to huge logs if the datasets had many (10k+) rows. 3) Compute training entropy after every epoch, not after every batch.By submitting this pull request, I confirm that my contribution is made under the terms of the Apache 2.0 license.