msaffari-amd ee3ada6e4a [AITERKER-112] PER_TOKEN_HEAD: support page_size < kN0 via cross-page dequant ...

- Pipeline: remove kPageBlockSize >= kN0 static_assert; QK dequant now
  precomputes tile_k_pages[] and indexes per-column. page_size >= kN0 stays
  on the original single-page fast path (kPagesPerTile==1).
- Codegen: add page_size=64 to SUPPORTED_PAGE_SIZE; drop per_token_head from
  the page_size < tile.F_bn0 filter (kv_blockscale still filtered).

2026-05-20 14:21:12 +00:00

..

__init__.py

chore(copyright): update copyright header for example directory (#3273)

2025-11-24 18:02:41 -08:00

fmha_batch_prefill.py

[AITERKER-112] PER_TOKEN_HEAD: support page_size < kN0 via cross-page dequant

2026-05-20 14:21:12 +00:00

fmha_bwd.py

[rocm-libraries] ROCm/rocm-libraries#7331 (commit 5692db0)

2026-05-14 13:34:32 +00:00

fmha_fwd_appendkv.py

[rocm-libraries] ROCm/rocm-libraries#4812 (commit bb5a4dd)

2026-02-25 16:13:56 +00:00

fmha_fwd_splitkv.py

[rocm-libraries] ROCm/rocm-libraries#6038 (commit d7041a2)

2026-04-03 00:18:21 +00:00

fmha_fwd.py

[rocm-libraries] ROCm/rocm-libraries#6529 (commit 93a6097)

2026-05-07 16:23:19 +00:00

fmha_pagedkv_prefill.py

[rocm-libraries] ROCm/rocm-libraries#5045 (commit 64a5502)

2026-03-03 21:55:14 +00:00