DeepSpeed
a54c3943 - [TiledFusedLogitsLoss] support inference (#7477)

Commit
139 days ago
[TiledFusedLogitsLoss] support inference (#7477) Adding inference support for `TiledFusedLogitsLoss` by skipping `backward` inside `forward` if the incoming tensor doesn't require grad. xref: https://github.com/snowflakedb/ArcticTraining/pull/259 --------- Signed-off-by: Stas Bekman <stas@stason.org> Co-authored-by: Rui Yan <49115835+yanrui27@users.noreply.github.com> Co-authored-by: Olatunji Ruwase <tunji.ruwase@snowflake.com>
Author
Parents
Loading