jgrusewski 9607f33518 perf(ml-alpha): block-per-row VSN bwd refactor (Phase B commit 3) ...

variable_selection_bwd refactored from grid=(1,1,1) to grid=(B*K,1,1).
VSN's n_rows = B*K positions (one row per (batch, K-position) pair);
block-per-row matches the existing fwd kernel's layout.

Adds 2 per-row grad scratch buffers + 2 reduce_axis0 launches:
  vsn_grad_w_scratch_d  [B*K, FEATURE_DIM, FEATURE_DIM]
  vsn_grad_b_scratch_d  [B*K, FEATURE_DIM]
~210 KB scratch at B=32, K=64.

VSN bwd runs 1×/step (not K×) so the absolute wall-time win here is
small versus commits 1+2. Done for pattern uniformity — every per-batch
or per-row bwd in the trainer now uses scratch+reducer.

All 9 perception_overfit smokes pass.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

2026-05-17 23:57:32 +02:00

8.6 KiB

Raw Blame History

View Raw