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
We found 2 specific examples of "dead code" where references appears to not be useful or relevant to mlperf.
Example (1): transformer running an extra eval on newstest 2013 (and printing eval acc).
Example (2): SSD saving snapshots/checkpoints and additional prints.
The text was updated successfully, but these errors were encountered:
We intend to eliminate "dead code" from the reference and submitters are not expected to implement this kind of dead code which is not relevant to the primary data pipeline/training/mlperf eval work.
Specific examples of what submitters don't need to implement:
Example (1): transformer running an extra eval on newstest 2013 (and printing eval acc).
Example (2): SSD saving snapshots/checkpoints and additional prints.
We note that we have already stated checkpoints are optional for MLPerf v0.6.
We found 2 specific examples of "dead code" where references appears to not be useful or relevant to mlperf.
Example (1): transformer running an extra eval on newstest 2013 (and printing eval acc).
Example (2): SSD saving snapshots/checkpoints and additional prints.
The text was updated successfully, but these errors were encountered: