[P/D disagg] Decode-side radix cache for SWA hybrid models (unified radix tree) (#27770)

Co-authored-by: Shangming Cai <csmthu@gmail.com>
This commit is contained in:
ishandhanani
2026-08-21 09:25:06 +08:00
committed by GitHub
co-authored by Shangming Cai
parent aa3f766799
commit 978244d671
10 changed files with 425 additions and 65 deletions
@@ -116,6 +116,7 @@ class TestDeepSeekV4HiSparseAllocator(CustomTestCase):
device=torch.device("cpu"),
page_size=256,
available_size=MagicMock(return_value=fill_len),
swa_available_size=MagicMock(return_value=swa_tail_len),
alloc_extend_swa_tail=MagicMock(return_value=kv_loc),
alloc_logical_only=MagicMock(return_value=kv_loc),
)