xla
[benchmarks] Fix execution with AMP precision.
#6512
Merged
Overview
Commits
6
Changes
View On GitHub

[benchmarks] Fix execution with AMP precision. #6512

ysiraichi merged 6 commits into master from ysiraichi/fix-amp
ysiraichi
ysiraichi Apply AMP on model forward only.
8a46a0fc
ysiraichi Fix import.
5d7d07ba
ysiraichi Add extra `device` argument for XLA AMP.
b6aa850d
ysiraichi Use `torch.amp.autocast` + assertion.
194350d8
ysiraichi Workaround for dynamo.
08a88544
ysiraichi Add path for inductor.
c901f076
ysiraichi ysiraichi added xla:gpu
ysiraichi ysiraichi requested a review from cota cota 2 years ago
ysiraichi ysiraichi requested a review from vanbasten23 vanbasten23 2 years ago
ysiraichi ysiraichi requested a review from frgossen frgossen 2 years ago
ysiraichi ysiraichi requested a review from golechwierowicz golechwierowicz 2 years ago
ysiraichi ysiraichi requested a review from zpcore zpcore 2 years ago
zpcore
ysiraichi
vanbasten23
ysiraichi
golechwierowicz
golechwierowicz approved these changes on 2024-02-12
ysiraichi ysiraichi merged 408b3766 into master 2 years ago

Login to write a write a comment.

Login via GitHub

Reviewers
Assignees
No one assigned
Labels
xla:gpu
Milestone
No milestone
Terms Of Use Privacy Policy Refund Policy Impressum