AlpinDale e14223dce5 kernel: use `cub::BlockReduce` instead of custom impl (#895) 3 weken geleden
..
amd f1d0b77c92 [0.6.0] Release Candidate (#481) 4 maanden geleden
nvidia f1d0b77c92 [0.6.0] Release Candidate (#481) 4 maanden geleden
common.cu e14223dce5 kernel: use `cub::BlockReduce` instead of custom impl (#895) 3 weken geleden
fp8_marlin.cu f1d0b77c92 [0.6.0] Release Candidate (#481) 4 maanden geleden