mirror of
https://github.com/ROCm/composable_kernel.git
synced 2026-05-04 13:41:24 +00:00
47a04fda08
[CK_TILE] Fix NaN for FMHA BWD When seq_q=0 ## Motivation This PR addresses NaNs in the FMHA backward (dQ/dK/dV) path when the effective query sequence length for a tile is zero, by ensuring the per-tile pipelines exit early with zeroed accumulators and by avoiding an early kernel return that prevented writing out cleared gradients. ## Technical Details - Add unconditional early-exit in the dK/dV pipelines when `num_total_loop <= 0` (no work), returning zeroed accumulators. - Adjust group-mode kernel early-return logic to only return when **both** `seqlen_q` and `seqlen_k` are zero, allowing blocks to run and store cleared dK/dV when `seqlen_q == 0`. ## Test Plan <!-- Explain any relevant testing done to verify this PR. --> ## Test Result <!-- Briefly summarize test outcomes. --> ## Submission Checklist - [x] Look over the contributing guidelines at https://github.com/ROCm/ROCm/blob/develop/CONTRIBUTING.md#pull-requests.