Skip to content

Commit 1dea708

Browse files
tingyangktingyangk
committed
Extend the BF16 fused grouped MLP to GeGLU, distributed weights, and FC2 scale_bias
Signed-off-by: tingyangk <tingyangk@nvidia.com>
1 parent 2192442 commit 1dea708

2 files changed

Lines changed: 866 additions & 449 deletions

File tree

0 commit comments

Comments
 (0)