SolitaryThinker f0b9c1773a [fix]: quantize fp8 weights lazily on the serving device (main converts post-GPU; CPU fp8 casts can round differently); ledger GateResult

2026-07-22 20:02:56 -07:00

..

__init__.py

[feat]: attention backend registry — FLASH_ATTN/SDPA, FASTVIDEO2_ATTENTION_BACKEND override, fail-closed selection, policy in env fingerprint

2026-07-22 16:16:42 -07:00

attention.py

[feat]: attention backend registry — FLASH_ATTN/SDPA, FASTVIDEO2_ATTENTION_BACKEND override, fail-closed selection, policy in env fingerprint

2026-07-22 16:16:42 -07:00

embeddings.py

[feat]: shared layers/ (norms, MLP, embeddings, rotary, attention) extracted from the Wan model — key- and cast-compatible

2026-07-22 15:42:14 -07:00

fp8.py

[fix]: quantize fp8 weights lazily on the serving device (main converts post-GPU; CPU fp8 casts can round differently); ledger GateResult

2026-07-22 20:02:56 -07:00

mlp.py

[feat]: shared layers/ (norms, MLP, embeddings, rotary, attention) extracted from the Wan model — key- and cast-compatible

2026-07-22 15:42:14 -07:00

norms.py

[feat]: shared layers/ (norms, MLP, embeddings, rotary, attention) extracted from the Wan model — key- and cast-compatible

2026-07-22 15:42:14 -07:00

rotary.py

[feat]: shared layers/ (norms, MLP, embeddings, rotary, attention) extracted from the Wan model — key- and cast-compatible

2026-07-22 15:42:14 -07:00

vsa.py

[feat]: FastWan VSA port — vendored tile math + kernel wrapper (fail-closed), VSA block/model, card; capture shim covers qad+vsa

2026-07-22 19:55:19 -07:00