Fairseq: What is wall and train_wall in the output log?

Created on 7 Feb 2019  路  2Comments  路  Source: pytorch/fairseq

In the scalling_nmt example training output log there is two column named "wall" and "train_wall".
What does these two column means? How this is related to training time?

Most helpful comment

"wall" is the total time, including time spent in validation, saving checkpoints to disk, etc. In the paper we reported "train_wall," which is just the time spent in training.

All 2 comments

"wall" is the total time, including time spent in validation, saving checkpoints to disk, etc. In the paper we reported "train_wall," which is just the time spent in training.

Thanks @myleott :)

Was this page helpful?
0 / 5 - 0 ratings

Related issues

jordiae picture jordiae  路  3Comments

mjpost picture mjpost  路  3Comments

tyoc213 picture tyoc213  路  3Comments

mali-nuist picture mali-nuist  路  3Comments

AranKomat picture AranKomat  路  3Comments