Download Latest Version
lmcache-0.3.7.tar.gz (1.1 MB)
Get an email when there's a new version of LMCache
| Name | Modified | Size | Downloads / Week |
|---|---|---|---|
| Parent folder | |||
| lmcache-0.3.5.tar.gz | 2025-08-29 | 1.0 MB | |
| lmcache-0.3.5-cp39-cp39-manylinux_2_24_x86_64.manylinux_2_28_x86_64.whl | 2025-08-29 | 3.8 MB | |
| lmcache-0.3.5-cp310-cp310-manylinux_2_24_x86_64.manylinux_2_28_x86_64.whl | 2025-08-29 | 3.8 MB | |
| lmcache-0.3.5-cp311-cp311-manylinux_2_24_x86_64.manylinux_2_28_x86_64.whl | 2025-08-29 | 3.8 MB | |
| lmcache-0.3.5-cp312-cp312-manylinux_2_24_x86_64.manylinux_2_28_x86_64.whl | 2025-08-29 | 3.8 MB | |
| lmcache-0.3.5-cp313-cp313-manylinux_2_24_x86_64.manylinux_2_28_x86_64.whl | 2025-08-29 | 3.8 MB | |
| README.md | 2025-08-28 | 3.0 kB | |
| v0.3.5 source code.tar.gz | 2025-08-28 | 995.7 kB | |
| v0.3.5 source code.zip | 2025-08-28 | 1.3 MB | |
| Totals: 9 Items | 22.3 MB | 0 | |
dear LMCache community, sincerest apologies of the unstable v0.3.4 release! It was caused the PR here: https://github.com/LMCache/LMCache/pull/1363.
v0.3.4 wheel and docker image (the lmcache/vllm-openai:latest) was unstable since vllm torch version was not upgraded yet to 2.8.0. We are now releasing v0.3.5 with pinned torch == 2.7.1 temporarily for a stable wheel + docker image. We are hoping to become:
1. more clear on which lmcache versions are compatible with which vllm versions (a compatibility matrix) soon
2. have our release system become more harmonious with vllm, especially as dependencies mismatch and kv connector interfaces change across future versions of vllm!
We will release an RFC on release management of LMCache and how we can try to stay well synced with the vllm ecosystem.
Some basic specs that we have in mind going forward: - do not break vllm's docker image when they import lmcache wheels (top priority) - keep dependencies synced with vllm (especially torch, which is used to compile and link binaries and breaks ABI compatibility across major versions) - allow for switching to different versions of vllm (and secondarily other serving engines) with lmcache from wheels, source, and docker image
What's Changed
- [CI/Build] Reduce pinned memory size in unit tests by @Shaoting-Feng in https://github.com/LMCache/LMCache/pull/1430
- [Bugfix] Compatible with latest vllm by @YaoJiayi in https://github.com/LMCache/LMCache/pull/1432
- [DOCS] Add a doc for internal_api_server by @maobaolong in https://github.com/LMCache/LMCache/pull/1419
- [patch]: busy loop on local cpu full by @sammshen in https://github.com/LMCache/LMCache/pull/1429
- [Core] Nixl storage backend by @tshmilnvidia in https://github.com/LMCache/LMCache/pull/1223
- [CI/Build] Reduce pinned memory in testing cache engine by @Shaoting-Feng in https://github.com/LMCache/LMCache/pull/1443
- [CI/Build] Harden linting by @YaoJiayi in https://github.com/LMCache/LMCache/pull/1448
- [Build]: Update Python lower and upper version support by @hickeyma in https://github.com/LMCache/LMCache/pull/1440
- [#1373][feat]Support dynamical load external backend by @maobaolong in https://github.com/LMCache/LMCache/pull/1439
- [CI][Linter] Add back strict mypy checking by @ApostaC in https://github.com/LMCache/LMCache/pull/1451
- [CI][Linter] Fix the mypy problem in CI by @ApostaC in https://github.com/LMCache/LMCache/pull/1452
- [feature] support heartbeat in lmcache worker by @chunxiaozheng in https://github.com/LMCache/LMCache/pull/652
- [Doc] update doc for nixl storage backend by @ziruiliu in https://github.com/LMCache/LMCache/pull/1450
New Contributors
- @tshmilnvidia made their first contribution in https://github.com/LMCache/LMCache/pull/1223
- @ziruiliu made their first contribution in https://github.com/LMCache/LMCache/pull/1450
Full Changelog: https://github.com/LMCache/LMCache/compare/v0.3.4...v0.3.5