Skip to content
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.

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

Set fp16 as default behavior if user has supported Nvidia GPU (or arm mac) #2203

Merged
merged 6 commits into from
Sep 13, 2023

Conversation

joeyballentine
Copy link
Member

Since we're resetting settings, I thought this would be pretty important to do. Whether or not a GPU can fp16 is determined by the arch enum kindly provided by the pynvml package. Now, users won't accidentally have fp16 off.

In the future, we could also do the same kind of thing for bf16 and implement support for that to speed up things like swinir on supported gpus.

backend/src/gpu.py Outdated Show resolved Hide resolved
backend/src/gpu.py Outdated Show resolved Hide resolved
backend/src/gpu.py Outdated Show resolved Hide resolved
backend/src/packages/chaiNNer_onnx/settings.py Outdated Show resolved Hide resolved
backend/src/packages/chaiNNer_pytorch/settings.py Outdated Show resolved Hide resolved
joeyballentine and others added 2 commits September 12, 2023 14:17
Co-authored-by: Michael Schmidt <mitchi5000.ms@googlemail.com>
backend/src/gpu.py Outdated Show resolved Hide resolved
backend/src/gpu.py Outdated Show resolved Hide resolved
backend/src/packages/chaiNNer_onnx/settings.py Outdated Show resolved Hide resolved
backend/src/packages/chaiNNer_pytorch/settings.py Outdated Show resolved Hide resolved
@joeyballentine joeyballentine merged commit 926ad07 into main Sep 13, 2023
14 checks passed
@joeyballentine joeyballentine deleted the fp16-default branch September 13, 2023 15:51
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

None yet

2 participants