Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Fix +/-inf in LSE returned by forward #978

Open
wants to merge 1 commit into
base: main
Choose a base branch
from
Open

Fix +/-inf in LSE returned by forward #978

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
12 changes: 6 additions & 6 deletions csrc/flash_attn/src/flash_fwd_kernel.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -100,7 +100,7 @@ inline __device__ void compute_attn_1rowblock(const Params &params, const int bi
#pragma unroll
for (int m = 0; m < size<1>(tOgO); ++m) {
const int row = get<0>(tOcO(0, m, 0));
if (row < binfo.actual_seqlen_q - m_block * kBlockM && get<1>(tOcO(0, m, 0)) == 0) { gLSE(row) = INFINITY; }
if (row < binfo.actual_seqlen_q - m_block * kBlockM && get<1>(tOcO(0, m, 0)) == 0) { gLSE(row) = -INFINITY; }
}
return;
}
Expand Down Expand Up @@ -545,7 +545,7 @@ inline __device__ void compute_attn_1rowblock_splitkv(const Params &params, cons
#pragma unroll
for (int m = 0; m < size<1>(tOgOaccum); ++m) {
const int row = get<0>(tOcO(0, m, 0));
if (row < binfo.actual_seqlen_q - m_block * kBlockM && get<1>(tOcO(0, m, 0)) == 0) { gLSEaccum(row) = Split ? -INFINITY : INFINITY; }
if (row < binfo.actual_seqlen_q - m_block * kBlockM && get<1>(tOcO(0, m, 0)) == 0) { gLSEaccum(row) = -INFINITY; }
}
return;
}
Expand Down Expand Up @@ -1141,17 +1141,17 @@ inline __device__ void combine_attn_seqk_parallel(const Params &params) {
for (int l = 1; l < kNLsePerThread; ++l) { lse_sum += expf(lse_accum(l) - lse_max); }
SumOp<float> sum_op;
lse_sum = Allreduce<kRowsPerLoadTranspose>::run(lse_sum, sum_op);
// For the case where all local lse == -INFINITY, we want to set lse_logsum to INFINITY. Otherwise
// lse_logsum is log(0.0) = -INFINITY and we get NaN when we do lse_accum(l) - lse_logsum.
ElementAccum lse_logsum = (lse_sum == 0.f || lse_sum != lse_sum) ? INFINITY : logf(lse_sum) + lse_max;
ElementAccum lse_logsum = logf(lse_sum) + lse_max;
// if (bidx == 0 && tidx < 32) { printf("tidx = %d, lse = %f, lse_max = %f, lse_logsum = %f\n", tidx, lse_accum(0), lse_max, lse_logsum); }
if (tidx % kRowsPerLoadTranspose == 0 && tidx / kRowsPerLoadTranspose < kBlockM) { gLSE(tidx / kRowsPerLoadTranspose) = lse_logsum; }
// Store the scales exp(lse - lse_logsum) in shared memory.
#pragma unroll
for (int l = 0; l < kNLsePerThread; ++l) {
const int row = l * kRowsPerLoadTranspose + tidx % kRowsPerLoadTranspose;
const int col = tidx / kRowsPerLoadTranspose;
if (row < params.num_splits && col < kBlockM) { sLSE[row][col] = expf(lse_accum(l) - lse_logsum); }
// For the case where all local lse == -INFINITY, we want to explicitly set sLSE to 0. Otherwise
// lse_logsum is log(0.0) = -INFINITY and we get NaN when we do lse_accum(l) - lse_logsum.
if (row < params.num_splits && col < kBlockM) { sLSE[row][col] = (lse_sum == 0.f || lse_sum != lse_sum) ? 0 : expf(lse_accum(l) - lse_logsum); }
}
__syncthreads();

Expand Down