Skip to content

Fixes CUDA 13 Linux wheel build jobs #4068

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 5 commits into from
Aug 28, 2025
Merged

Fixes CUDA 13 Linux wheel build jobs #4068

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
fe966ad
fix??
NicolasHug Aug 28, 2025
5617d10
Use cuda::std::numeric_limits
NicolasHug Aug 28, 2025
6bbd50e
fix forced_align?
NicolasHug Aug 28, 2025
6fdfbf4
Fix for 12.6 and 12.8?
NicolasHug Aug 28, 2025
83b54e9
Use CUDART_VERSION instead of CUDA_VERSION
NicolasHug Aug 28, 2025
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
5 changes: 5 additions & 0 deletions src/libtorchaudio/cuctc/src/ctc_prefix_decoder_kernel_v2.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -25,6 +25,7 @@
// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
#include <float.h>
#include <algorithm>
#include <limits>
#include "../include/ctc_prefix_decoder_host.h"
#include "ctc_fast_divmod.cuh"
#include "cub/cub.cuh"
Expand Down Expand Up @@ -440,7 +441,11 @@ __launch_bounds__(BLOCK_SIZE) void topk_reduce_and_copy_list_per_batch_kernel(
topk_values,
beam,
items_per_batch,
#if CUDART_VERSION >= 12090 // CUDA 12.9 and later
std::numeric_limits<float>::lowest(),
#else
cub::FpLimits<float>::Lowest(),
#endif
block_topk_fun,
set_key_value);

Expand Down
4 changes: 4 additions & 0 deletions src/libtorchaudio/forced_align/gpu/compute.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -94,7 +94,11 @@ __global__ void falign_cuda_step_kernel(
alphas_a[curIdxOffset][i] = result + logProbs_a[batchIndex][t][labelIdx];
threadMax = max(threadMax, alphas_a[curIdxOffset][i]);
}
#if CUDART_VERSION >= 12090 // CUDA 12.9 and later
scalar_t maxResult = BlockReduce(tempStorage).Reduce(threadMax, thrust::maximum<scalar_t>());
#else
scalar_t maxResult = BlockReduce(tempStorage).Reduce(threadMax, cub::Max());
#endif
if (threadIdx.x == 0) {
maxValue = maxResult;
}
Expand Down
Loading