Skip to content

Fix cuDNN convolution precision on Ampere+ GPUs #3127

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
davisking merged 1 commit into from
Dec 28, 2025
+9 −0
Merged

Fix cuDNN convolution precision on Ampere+ GPUs #3127

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
9 changes: 9 additions & 0 deletions dlib/cuda/cudnn_dlibapi.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1044,6 +1044,15 @@ namespace dlib
CUDNN_CROSS_CORRELATION)); // could also be CUDNN_CONVOLUTION
#endif

#if CUDNN_MAJOR >= 8
// On Ampere and later GPUs, CUDNN_DEFAULT_MATH permits TF32 Tensor Core
// operations which have reduced precision. Use CUDNN_FMA_MATH to force
// true FP32 computation for consistent numerical results.
CHECK_CUDNN(cudnnSetConvolutionMathType(
(cudnnConvolutionDescriptor_t)conv_handle,
CUDNN_FMA_MATH));
#endif

CHECK_CUDNN(cudnnGetConvolution2dForwardOutputDim(
(const cudnnConvolutionDescriptor_t)conv_handle,
descriptor(data),
Expand Down
Loading