Skip to content

Commit

Permalink
Add missing schedules argument in SM90 fp16 op generation (#920)
Browse files Browse the repository at this point in the history
  • Loading branch information
aakhundov authored and ttl10101 committed Feb 7, 2024
1 parent 3646018 commit 8833890
Showing 1 changed file with 1 addition and 1 deletion.
2 changes: 1 addition & 1 deletion tools/library/scripts/generator.py
Original file line number Diff line number Diff line change
Expand Up @@ -4160,7 +4160,7 @@ def GenerateSM90_TensorOp_16b_WGMMA_gemm(manifest, cuda_version):
elif data_type_mixed["c_type"] in [DataType.f16, DataType.bf16]:
layout[2][1] = 8

CreateGemmUniversal3xOperator(manifest, layouts, tile_descriptions, data_type_mixed)
CreateGemmUniversal3xOperator(manifest, layouts, tile_descriptions, data_type_mixed, schedules)
# persistent kernels with TMA epilogues
if data_type_mixed["c_type"] in [DataType.f16, DataType.bf16] and CudaToolkitVersionSatisfies(cuda_version, 12, 1):
CreateGemmUniversal3xOperator(manifest, layouts, tile_descriptions, data_type_mixed,
Expand Down

0 comments on commit 8833890

Please sign in to comment.