A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 14, 05:30 UTC
[UT] fix unify_kv_cache_configs when kv cache config needs sort (#23843)
Ning Xie committed
Aug 30, 2025 at 19:22 UTC
5490d633cec0e6b946d3f5c2d56e6236ef42eb40
1 file changed
+8
-2
tests/v1/core/test_kv_cache_utils.py
+8
-2
@@ -601,8 +601,14 @@ def test_unify_kv_cache_configs():
601
]
602
603
unify_kv_cache_configs(need_sort_kv_cache_config)
604
- assert need_sort_kv_cache_config[0].num_blocks == 10
605
- assert need_sort_kv_cache_config[1].num_blocks == 10
604
+ sorted_kv_cache_groups = [
605
+ KVCacheGroupSpec(["layer1"], new_kv_cache_spec()),
606
+ KVCacheGroupSpec(["layer2"], new_kv_cache_spec(num_kv_heads=4)),
607
+ ]
608
+ assert (
609
+ need_sort_kv_cache_config[0].kv_cache_groups == sorted_kv_cache_groups)
610
+ assert (
611
+ need_sort_kv_cache_config[1].kv_cache_groups == sorted_kv_cache_groups)
612
613
diff_kv_cache_config = [
614
KVCacheConfig(