ViperEkura a29bdfae46 refactor: rework attention backend resolution ...

- explicit attn_backend() context wins over ASTR_BACKEND env
- polymorphic available()/supports_call() replace isinstance dispatch
- cache singleton backend instances to avoid hot-path allocation
- training (fwd=None) resolves cuda > flash > torch by capability
- flash dense supports mask-free calls only; masked training falls back to torch

2026-08-23 14:47:02 +08:00

..

developer

refactor: rework attention backend resolution

2026-08-23 14:47:02 +08:00

guides

refactor: rework attention backend resolution

2026-08-23 14:47:02 +08:00

images

docs: restructure to docs/, add guides and developer docs

2026-07-30 00:49:04 +08:00

get-started.md

docs: fix stale docs and align with code

2026-08-23 14:23:28 +08:00

README-zh-CN.md

docs: fix stale docs and align with code

2026-08-23 14:23:28 +08:00