ViperEkura 05739629fc feat: add timed() context manager and backend supports() ...

- Each backend exposes static supports(**kwargs) for capability query
- CudaBackend.supports checks head_dim + kernel availability
- FlashAttnBackend/TorchNativeBackend always return True
- timed() context manager gated by ASTRAI_TIMED=1 env var, logs via logger.info
- Wraps warmup prefill/decode, execute_prefill, and execute_decode

2026-08-07 20:51:30 +08:00

..

lib

refactor: move compiled kernel .so files into extension/lib

2026-07-31 15:36:32 +08:00

__init__.py

refactor : use factory for attention backends

2026-08-05 15:37:22 +08:00

attention_backend.py

feat: add timed() context manager and backend supports()

2026-08-07 20:51:30 +08:00

attention_ops.py

fix: add out_buf to attn_paged_decode for CUDA graph capture compatibility

2026-08-07 19:45:59 +08:00

loader.py

feat: SGLang-style paged attention kernels replace page-table path

2026-08-01 15:41:25 +08:00

rotary_backend.py

refactor: unify rotary embedding interface and update docs

2026-07-31 16:52:25 +08:00

rotary_ops.py

refactor: unify rotary embedding interface and update docs

2026-07-31 16:52:25 +08:00