Skip to content

update 0.3.7-dev#120

Closed
chloroethylene wants to merge 23 commits into
v0.3.7-devfrom
main
Closed

update 0.3.7-dev#120
chloroethylene wants to merge 23 commits into
v0.3.7-devfrom
main

Conversation

@chloroethylene

Copy link
Copy Markdown
Collaborator

update 0.3.7-dev

shengyan-chen and others added 23 commits December 2, 2025 10:35
* New kvcache format adapted for vLLM 0.11.0
* Feature: Support for Online RAG Benchmarking on LMCache v1

* CacheBlend support for v0.3.7

* Add support files

* fix bug:AttributeError: 'StorageManager' object has no attribute 'async_serializer'

* Update rag benchmark

* .

* ..

* [Doc]:Some ad-hoc changes needed in vLLM-Ascend for CacheBlend Integration

* [DOC]:update readme.txt

* .....

* :

---------

Co-authored-by: sunshine <xianiming1@huawei.com>
* Add docker ci build sh

* Update readme to include compat matrix, and added example

* Update Dockerfile for new base image and dependencies

---------

Signed-off-by: Matthew Yeung <yyygggfff@hotmail.com>
Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
Co-authored-by: ehuohz <zhouhe2025@gmail.com>
* Fix Readme with wrong column title

Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
Signed-off-by: Matthew Yeung <yyygggfff@hotmail.com>

---------

Signed-off-by: Matthew Yeung <yyygggfff@hotmail.com>
Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
* [refactor] remove host_register as it is not used anymore
* bug fix: 310p ms

* fix bug:310p registration

* .
* [bugfix] fix mindspore numpy tensor bug
…97)

* Fix layerwise offload in sync mode

---------

Signed-off-by: Matthew Yeung <yyygggfff@hotmail.com>
add an FAQ for lack of `numaif.h` header file

Signed-off-by: Junyuan <48611576+chloroethylene@users.noreply.github.com>
Signed-off-by: liwei <arch66@qq.com>
)

- Implement a test bootstrap system in `conftest.py` that automatically
  clones/syncs upstream LMCache at a specific version tag.
- Register the upstream `tests` folder as a module alias (`lmcache_tests`)
  to allow direct imports.
- Refactor existing tests to use the alias, removing hundreds of lines of 
  duplicated code.
- Apply runtime patching to swap GPU connectors with NPU-specific classes
  (using `unittest.mock.patch` with `new=...`).

Signed-off-by: Matthew Yeung <yyygggfff@hotmail.com>
Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
* Update Ascend Driver version in README

Signed-off-by: Junyuan <48611576+chloroethylene@users.noreply.github.com>

* Update CANN Toolkit version format in README

Signed-off-by: Junyuan <48611576+chloroethylene@users.noreply.github.com>

---------

Signed-off-by: Junyuan <48611576+chloroethylene@users.noreply.github.com>
* Enhance error message for aclrtHostRegister failure

Added a driver version check in error message for aclrtHostRegister failure.

Signed-off-by: Junyuan <48611576+chloroethylene@users.noreply.github.com>

* Simplify error messages in managed_mem.cpp

Removed driver version check message from error logs for aclrtHostRegister and halHostRegister.

Signed-off-by: Junyuan <48611576+chloroethylene@users.noreply.github.com>

* Update managed_mem.cpp

Signed-off-by: Junyuan <48611576+chloroethylene@users.noreply.github.com>

---------

Signed-off-by: Junyuan <48611576+chloroethylene@users.noreply.github.com>
* Add build workflow check

Signed-off-by: Matthew Yeung <yyygggfff@hotmail.com>

* Update Dockerfile paths in build workflow

Signed-off-by: Matthew Yeung <yyygggfff@hotmail.com>

* Enable submodules in build workflow

Signed-off-by: Matthew Yeung <yyygggfff@hotmail.com>

* Add Pre-Commit & GitHub actions (#1)

* Add Pre-commit lint & build test workflow action
---------

Signed-off-by: Matthew Yeung <yyygggfff@hotmail.com>

* Update build-and-test.yml

- because the host-ip variable cannot work here yet, so temporarily pass for the bridgeip

Signed-off-by: Matthew Yeung <yyygggfff@hotmail.com>

* Add pull-requests permission to workflow

Signed-off-by: Matthew Yeung <yyygggfff@hotmail.com>

* Remove test report publishing step

Removed the test report publishing step from the workflow.

Signed-off-by: Matthew Yeung <yyygggfff@hotmail.com>

* Add workflow to publish test results

Signed-off-by: Matthew Yeung <yyygggfff@hotmail.com>

* enhance reports xml

* Fix formatting in build-and-test workflow

Signed-off-by: Matthew Yeung <yyygggfff@hotmail.com>

* Fix formatting in report-test-results.yml

Signed-off-by: Matthew Yeung <yyygggfff@hotmail.com>

* Fix error message formatting in managed_mem.cpp

Signed-off-by: Matthew Yeung <yyygggfff@hotmail.com>

---------

Signed-off-by: Matthew Yeung <yyygggfff@hotmail.com>
* Update issue templates

* Enhance bug report template with logs section

Added a section for logs in the bug report template.

Signed-off-by: Matthew Yeung <yyygggfff@hotmail.com>

* Enhance feature request template with design details

Added guidance for providing preliminary code design and planned modifications.

Signed-off-by: Matthew Yeung <yyygggfff@hotmail.com>

* Update .github/ISSUE_TEMPLATE/bug_report.md

Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
Signed-off-by: Junyuan <48611576+chloroethylene@users.noreply.github.com>

---------

Signed-off-by: Matthew Yeung <yyygggfff@hotmail.com>
Signed-off-by: Junyuan <48611576+chloroethylene@users.noreply.github.com>
Co-authored-by: Junyuan <48611576+chloroethylene@users.noreply.github.com>
Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
Removed unnecessary std::endl statement after error message.
…en using MLA (#110)

* fix: Fix KVCache format detection when using MLA

* Implementing Staging Copy C++ Operator instead of ZeroCopy / NativeTorch copy + kernel transfer
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

5 participants