biondizzle ca5bc814d5 Fix compressor: do not add positional bias to KV content ...

The positional bias (ape/B) should only modulate the compression
softmax logits (Z + B), NOT be added to the KV content itself.

Paper equation: compressed = softmax(Z + B) · C
Bug was doing: compressed = softmax(Z + B) · (C + B) — poisons every
compressed KV entry with learned positional-bias content.

Fixed in both CSA (compress_csa_reduce_kernel) and HCA
(hca_compress_reduce_kernel) paths in compressor_reduce.cu.

2026-06-03 15:52:00 +00:00

..

attention

Wire prefill FMHA into production.py and single_shot

2026-06-03 03:49:57 +00:00

cache

Cleanup Step 2: Archive Lineage P code, fix broken imports

2026-06-02 19:27:07 +00:00

compressor

Cleanup Step 2: Archive Lineage P code, fix broken imports

2026-06-02 19:27:07 +00:00

cuda

Fix compressor: do not add positional bias to KV content

2026-06-03 15:52:00 +00:00

gemm

fix: use cute.where() directly for clamp in fused SwiGLU

2026-06-02 08:16:41 +00:00

indexer

Cleanup Step 2: Archive Lineage P code, fix broken imports

2026-06-02 19:27:07 +00:00

router

Cleanup Step 2: Archive Lineage P code, fix broken imports

2026-06-02 19:27:07 +00:00

__init__.py

Restructure: cutedsl/ -> dsv4/ with proper layering

2026-05-21 17:30:44 +00:00