github/ColossalAI
1
0
Fork 0
mirror of https://github.com/hpcaitech/ColossalAI.git synced 2025-08-27 20:22:09 +00:00
Code Issues Packages Projects Releases Wiki Activity

fix format (#564)

Browse Source
This commit is contained in:
coder-chin 2022-03-31 15:00:50 +08:00 committed by binmakeswell
parent e014144c44
commit 5835631218
1 changed files with 4 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

3
colossalai/kernel/cuda_native/csrc/kernels/softmax_kernels.cu
View File

@ -304,7 +304,8 @@ __global__ void ker_attn_softmax_bw(T *grad, const T *inp, int softmax_length) {
cg::thread_block b = cg::this_thread_block(); cg::thread_block b = cg::this_thread_block();
cg::thread_block_tile<WARP_SIZE> g = cg::tiled_partition<WARP_SIZE>(b); cg::thread_block_tile<WARP_SIZE> g = cg::tiled_partition<WARP_SIZE>(b);
for (int i = 1; i < WARP_SIZE; i <<= 1) sum += g.shfl_xor(sum, i); for (int i = 1; i < WARP_SIZE; i <<= 1)
sum += g.shfl_xor(sum, i);
#pragma unroll #pragma unroll
for (int i = 0; i < ITERATIONS; ++i) { for (int i = 0; i < ITERATIONS; ++i) {