projects / platform / upstream / pytorch.git / commit
? search:
summary | shortlog | log | commit | commitdiff | tree
(parent: 93f1090) | patch
add operation list for AutocastCPU (#63534)
authorleslie-fang-intel <leslie.fang@intel.com>
Tue, 31 Aug 2021 02:28:59 +0000 (19:28 -0700)
committerFacebook GitHub Bot <facebook-github-bot@users.noreply.github.com>
Tue, 31 Aug 2021 02:30:33 +0000 (19:30 -0700)
commit09dfaa0339c944d1c4b64193e0962972affbe0c7
tree922ac1abf96c949754d5a126581a6ac39113747ftree | snapshot
parent93f1090267df7e3023017b83f885edbc59e48913commit | diff
add operation list for AutocastCPU (#63534)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/63534

In this PR:
* We have changed the default dtype of `AutocastCPU` from `float16` to `bfloat16` as discussed here `https://github.com/pytorch/pytorch/pull/61002`
* We also update the operation list which needs casting to `lower_precision_fp` or `float32`.

Test Plan: Imported from OSS

Reviewed By: zou3519

Differential Revision: D30644914

Pulled By: ezyang

fbshipit-source-id: 8b93485ba452b3759611e3f0ac88e920fe495ac1
aten/src/ATen/autocast_mode.cpp diff | blob | history
test/run_test.py diff | blob | history
torch/cpu/amp/autocast_mode.py diff | blob | history
torch/testing/_internal/autocast_test_lists.py diff | blob | history
Domain: Machine Learning / ML Framework;