ViperEkura 3fa7e66676 refactor: decouple task cache from PagePool and unify steady-state detection ...

- TaskCacheRegistry -> TaskCacheManager (independent, held by scheduler)
- TaskCacheState co-locates 5 parallel dicts into one dataclass
- AllocationStrategy base class + PagedStrategy subclass (page_size is a parameter)
- _rollback() helper for unified cleanup (no duplicate free paths)
- Task._kv_len + prefill_done property (explicit, no output_tokens proxy)
- Steady-state detection single-sourced in TaskCacheManager.bind()
- PagePool is now pure physical layer (no task knowledge)
- Removed dead _page_to_hash dict in RadixCache

2026-08-08 22:43:06 +08:00

..

conftest.py

refactor : FastAPI 懒加载单例，消除模块级副作用

2026-06-04 15:52:27 +08:00

test_cache.py

refactor: decouple task cache from PagePool and unify steady-state detection

2026-08-08 22:43:06 +08:00

test_engine.py

fix : handle zero-token batch generation

2026-08-06 12:31:09 +08:00

test_protocol.py

refactor: simplify inference engine and backend dispatch

2026-08-07 22:28:48 +08:00

test_sample.py

refactor: unify rollout onto inference engine KV-cache path

2026-07-20 12:52:20 +08:00

test_scheduler.py

fix: add dtype guard to FlashAttnBackend capability check

2026-08-07 23:08:22 +08:00

test_server.py

refactor : FastAPI 懒加载单例，消除模块级副作用

2026-06-04 15:52:27 +08:00

test_task.py

refactor: decouple task cache from PagePool and unify steady-state detection

2026-08-08 22:43:06 +08:00

test_tool_parser.py

test: prune low-value and duplicate tests

2026-08-01 16:01:20 +08:00