minke.yu/sglang
1
0
Fork 0
Code Issues Pull Requests Actions 12 Packages Projects Releases Wiki Activity

[FEAT] Decouple multimodal global cache from Mooncake (#30392)

Browse Source 
Co-authored-by: Yuang Chen <cya539102@antgroup.com>
Co-authored-by: Yuang Chen <1131578721@qq.com>
This commit is contained in:
siyu
2026-08-10 19:23:19 +08:00
committed by GitHub
co-authored by Yuang Chen Yuang Chen
parent d96b1533ea
commit 14ffd447a4
6 changed files with 153 additions and 12 deletions
Download Patch File Download Diff File Expand all files Collapse all files
test/registered/unit/mem_cache/test_embedding_cache_controller.py
+2 -2
View File
@@ -8,7 +8,7 @@ from unittest.mock import MagicMock
import torch
from sglang.srt.managers.schedule_batch import Modality
from sglang.srt.mem_cache.storage.mooncake_store.embedding_cache_controller import (
from sglang.srt.mem_cache.embedding_cache_controller import (
EmbeddingCacheController,
EmbeddingCacheEntry,
EmbeddingPool,
@@ -55,7 +55,7 @@ def _make_controller(num_pages=16, dim=4, page_size=2, enable_eviction=True):
ctrl.element_size = torch.float32.itemsize
ctrl.enable_eviction = enable_eviction
ctrl.max_eviction_batch = 10
ctrl.mooncake_store = MagicMock()
ctrl.embedding_store = MagicMock()
ctrl.total_pool_size_bytes = num_pages * page_size * dim * torch.float32.itemsize
ctrl.vision_pool = _make_pool(num_pages, dim, page_size)
ctrl.audio_pool = _make_pool(num_pages, dim, page_size, modality="audio")