Skip to content

added bf16 support #25

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 1 commit into from
Sep 8, 2023
Merged

added bf16 support #25

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
22 changes: 1 addition & 21 deletions src/transformers/utils/import_utils.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -264,27 +264,7 @@ def is_torch_bf16_gpu_available():
return False

import torch

# since currently no utility function is available we build our own.
# some bits come from https://github.com/pytorch/pytorch/blob/2289a12f21c54da93bf5d696e3f9aea83dd9c10d/torch/testing/_internal/common_cuda.py#L51
# with additional check for torch version
# to succeed: (torch is required to be >= 1.10 anyway)
# 1. the hardware needs to support bf16 (GPU arch >= Ampere, or CPU)
# 2. if using gpu, CUDA >= 11
# 3. torch.autocast exists
# XXX: one problem here is that it may give invalid results on mixed gpus setup, so it's
# really only correct for the 0th gpu (or currently set default device if different from 0)
if torch.cuda.is_available() and torch.version.cuda is not None:
if torch.cuda.get_device_properties(torch.cuda.current_device()).major < 8:
return False
if int(torch.version.cuda.split(".")[0]) < 11:
return False
if not hasattr(torch.cuda.amp, "autocast"):
return False
else:
return False

return True
return torch.cuda.is_bf16_supported()


def is_torch_bf16_cpu_available():
Expand Down