Skip to content

Commit

Permalink
Merge branch 'mamba_flops' of https://github.com/EleutherAI/gpt-neox
Browse files Browse the repository at this point in the history
…into mamba_flops
  • Loading branch information
Quentin-Anthony committed Sep 23, 2024
2 parents 4ad542a + b2c1b01 commit d932b58
Showing 1 changed file with 2 additions and 2 deletions.
4 changes: 2 additions & 2 deletions megatron/fused_kernels/type_shim.h
Original file line number Diff line number Diff line change
Expand Up @@ -277,7 +277,7 @@ reduce_block_into_lanes(T* x,
final = x[tid] + x[tid + 32];
else
final = val;
// __SYNCWARP();
// __SYNCWARP();

#pragma unroll
for (int i = 16; i >= lanes; i >>= 1)
Expand Down Expand Up @@ -321,7 +321,7 @@ reduce_block_into_lanes_max_op(T* x,
final = fmaxf(fabsf(x[tid]), fabsf(x[tid + 32]));
else
final = val;
// __SYNCWARP();
// __SYNCWARP();

#pragma unroll
for (int i = 16; i >= lanes; i >>= 1)
Expand Down

0 comments on commit d932b58

Please sign in to comment.