fix: isolate continuous batch decode state

- Match steady-state metadata to the active task IDs - Rebuild request mappings for cached prefix pages - Add regressions for batch refill and prefix reuse
2026-08-09 01:01:41 +08:00
parent a33ca04f60
commit be90dfe2bd
4 changed files with 78 additions and 3 deletions
@@ -299,7 +299,7 @@ class PagedStrategy(AllocationStrategy):

    def write_indices(self, state: TaskCacheState, prompt_ids: List[int]) -> None:
        total = len(prompt_ids)
-        for pos in range(state.cached, total):
+        for pos in range(total):
            page_idx = pos // self._page_size
            offset = pos % self._page_size
            if page_idx < len(state.pages):