[WIP] Before refactor policies.

2026-01-06 20:47:55 +08:00
parent 7cc8a394a5
commit 690492e074
6 changed files with 112 additions and 237 deletions
--- a/nanovllm/kvcache/sparse/full_policy.py
+++ b/nanovllm/kvcache/sparse/full_policy.py
@@ -22,6 +22,10 @@ class FullAttentionPolicy(SparsePolicy):
    - For short sequences where sparsity isn't beneficial
    """

+    # Full attention supports both prefill and decode
+    supports_prefill = True
+    supports_decode = True
+
    def select_blocks(
        self,
        available_blocks: List[int],