[Feature] OFFLOAD: standalone LMCache CPU/NVMe KV-offload connector - #1318

Merged

valarLip merged 11 commits into

yhl-amd:feature/lmcache-offload-merge

Jun 29, 2026

Commits on Jun 29, 2026

[Feature] OFFLOAD: add LMCache CPU/NVMe KV-offload subsystem

yhl-amd
and
claude
committed
b060da0
[Feature] OFFLOAD: integrate KV offload load/save into engine and scheduler

yhl-amd
and
claude
committed
26ca242
[Frontend] Support max completion tokens in OpenAI API

yhl-amd
and
claude
committed
9aaff8e
[Docs] Add design README for LMCache CPU/NVMe KV offload connector

yhl-amd
and
claude
committed
1d175a0
[Feature] OFFLOAD: support MLA models (DeepSeek R1/V3, Kimi)

yhl-amd
and
claude
committed
4bd333f
[Docs] Update LMCache offload README for MLA support

yhl-amd
and
claude
committed
d52ba0b
[Enhancement] OFFLOAD: make unaligned handoff always-on; drop OFFLOAD_UNALIGNED_HANDOFF

yhl-amd
and
claude
committed
846916c
[Bugfix] OFFLOAD: drop duplicate enable_chunked_prefill kwarg in test MockConfig

yhl-amd
and
claude
committed
61b2f84
[Bugfix] OFFLOAD: drop stale unaligned-skip test left over from always-on handoff

yhl-amd
and
claude
committed
30d9f59
[OFFLOAD] Remove Chinese README.zh-CN.md

yhl-amd
and
claude
committed
ac394b6
Merge branch 'main' into feature/lmcache-offload-merge
yhl-amd
authored
2356fb9