search

mlcommons / training_policies

Issues related to MLPerf™ training policies, including rules and suggested changes
https://mlcommons.org/en/groups/training
Apache License 2.0
92 stars 66 forks source link

DLRM - remove unnecessary eval metrics in submission #321

Closed christ1ne closed 4 years ago

christ1ne commented 4 years ago

Hello all,

The DLRM code currently has multiple metrics in the eval step. I think per the training rules we can remove the unnecessary ones and only keep the one per the target accuracy requirement.

just want to confirm in the WG.

bitfort commented 4 years ago

SWG:

There are multiple metrics computed during eval for DLRM in the reference, we clarify that the only metric that needs to be computed are the metrics directly needed for the stopping criteria.

bitfort commented 4 years ago

Make this documented in the rules.

petermattson commented 4 years ago

Fixed per PR above.