biondizzle 7f69979c5f D1.5: add multi-KV-tile attention test with Python KV merge ...

- Splits K/V into 128-token segments
- Runs FMHA per segment, merges with exp(lse) weighted sum
- Tests: s_k=256 (2 tiles), s_k=512 (4 tiles)
- Uses reference attn_sum for normalization

2026-05-25 17:18:50 +00:00

7.0 KiB

Raw Blame History

View Raw