Skip to content

update turbomind builder image - #4752

Merged
lvhan028 merged 18 commits into
InternLM:mainfrom
irexyc:manylinux_2_28
Jul 17, 2026
Merged

update turbomind builder image#4752
lvhan028 merged 18 commits into
InternLM:mainfrom
irexyc:manylinux_2_28

Conversation

@irexyc

@irexyc irexyc commented Jul 15, 2026

Copy link
Copy Markdown
Collaborator

Thanks for your contribution and we appreciate it a lot. The following instructions would make your pull request more healthy and more easily receiving feedbacks. If you do not understand some items, don't worry, just make the pull request and seek help from maintainers.

Motivation

Please describe the motivation of this PR and the goal you want to achieve through this PR.

Modification

Please briefly describe what modification is made in this PR.

BC-breaking (Optional)

Does the modification introduce changes that break the backward-compatibility of the downstream repositories?
If so, please describe how it breaks the compatibility and how the downstream projects should modify their code to keep compatibility with this PR.

Use cases (Optional)

If this PR introduces a new feature, it is better to list some use cases here, and update the documentation.

Checklist

  1. Pre-commit or other linting tools are used to fix the potential lint issues.
  2. The modification is covered by complete unit tests. If not, please add more unit tests to ensure the correctness.
  3. If the modification has a dependency on downstream projects of a newer version, this PR should be tested with all supported versions of downstream projects.
  4. The documentation has been modified accordingly, like docstring or example tutorials.

@irexyc
irexyc marked this pull request as ready for review July 15, 2026 04:07
Copilot AI review requested due to automatic review settings July 15, 2026 04:07

Copilot AI left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Pull request overview

Updates the manywheel/turbomind builder workflow to use a manylinux_2_28-based image and aligns builder/wheel build scripts with the new tagging and Dockerfile naming, while also removing a hard dependency on the GNU gold linker from the CMake build.

Changes:

  • Remove forced ld.gold usage from CMakeLists.txt so builds can use the default linker.
  • Add a manylinux_2_28 builder Dockerfile and update builder image tagging to include the manylinux version.
  • Adjust build scripts to default toward manylinux_2_28 (builder + wheel builds).

Reviewed changes

Copilot reviewed 5 out of 5 changed files in this pull request and generated 3 comments.

Show a summary per file
File Description
CMakeLists.txt Removes the enforced gold-linker requirement/flags.
builder/manywheel/Dockerfile_2_28 Adds a manylinux_2_28-based builder image definition (CUDA + conda envs).
builder/manywheel/build_lmdeploy_builder.sh Updates builder image tag/Dockerfile selection for manylinux-versioned builds.
builder/manywheel/build_all_wheel.sh Switches default platform to manylinux_2_28 and updates wheel build tag naming.
builder/manywheel/build_all_lmdeploy_builders.sh Makes manylinux version and CUDA version configurable via env defaults.

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

Comment thread builder/manywheel/Dockerfile_2_28 Outdated
Comment thread builder/manywheel/build_lmdeploy_builder.sh
Comment thread builder/manywheel/build_all_wheel.sh
Comment thread builder/manywheel/scripts/install_cuda.sh
Comment thread builder/manywheel/build_all_wheel.sh Outdated
Comment thread builder/manywheel/build_lmdeploy_builder.sh
Comment thread builder/manywheel/Dockerfile_2_28 Outdated
@lvhan028

Copy link
Copy Markdown
Collaborator

Please fix the linting error

Comment thread docs/zh_cn/inference/turbomind.md Outdated
Comment thread docs/en/inference/turbomind.md Outdated

Copilot AI left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Pull request overview

Copilot reviewed 22 out of 22 changed files in this pull request and generated 14 comments.

Comment thread docs/en/inference/turbomind.md Outdated
Comment thread docs/zh_cn/inference/turbomind.md Outdated
Comment thread builder/manywheel/scripts/install_cuda.sh Outdated
Comment thread builder/manywheel/scripts/install_cuda.sh Outdated
Comment thread builder/manywheel/build_all_wheel.sh Outdated
Comment thread .github/workflows/benchmark.yml Outdated
Comment thread .github/workflows/daily_ete_test.yml
Comment thread .github/workflows/daily_ete_test_5080.yml Outdated
Comment thread .github/workflows/daily_ete_test_3090.yml Outdated
Comment thread .github/workflows/mllm_api_eval.yml Outdated
Comment thread .github/workflows/stable.yml Outdated

Copilot AI left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Pull request overview

Copilot reviewed 22 out of 22 changed files in this pull request and generated 4 comments.

Comment on lines +110 to +126
function install_130 {
echo "Installing CUDA 13.0 and NCCL 2.30.3"
rm -rf /usr/local/cuda-13.0 /usr/local/cuda
# install CUDA 13.0.3 in the same container
wget -q https://developer.download.nvidia.com/compute/cuda/13.0.3/local_installers/cuda_13.0.3_580.126.20_linux.run
chmod +x cuda_13.0.3_580.126.20_linux.run
./cuda_13.0.3_580.126.20_linux.run --toolkit --silent
rm -f cuda_13.0.3_580.126.20_linux.run
rm -f /usr/local/cuda && ln -s /usr/local/cuda-13.0 /usr/local/cuda

# NCCL license: https://docs.nvidia.com/deeplearning/nccl/#licenses
mkdir tmp_nccl && cd tmp_nccl
wget -q https://developer.download.nvidia.com/compute/redist/nccl/v2.30.3/nccl_2.30.3-1+cuda13.2_x86_64.txz
tar xf nccl_2.30.3-1+cuda13.2_x86_64.txz
cp -a nccl_2.30.3-1+cuda13.2_x86_64/include/* /usr/local/cuda/include/
cp -a nccl_2.30.3-1+cuda13.2_x86_64/lib/* /usr/local/cuda/lib64/
cd ..
Comment on lines 25 to 29
docker_build_args=(
-t "${DOCKER_IMAGE}"
--target "${TARGET}"
-f "${TOPDIR}/manywheel/Dockerfile${DOCKERFILE_SUFFIX}"
-f "${TOPDIR}/manywheel/Dockerfile_${MANY_LINUX_VERSION}"
)
Comment thread builder/windows/setup_cuda.ps1
Comment thread .github/workflows/linux_x64_gpu.yml

Copilot AI left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Pull request overview

Copilot reviewed 22 out of 22 changed files in this pull request and generated 2 comments.


# NCCL license: https://docs.nvidia.com/deeplearning/nccl/#licenses
mkdir tmp_nccl && cd tmp_nccl
wget -q https://developer.nvidia.com/w/compute/redist/nccl/v2.28.9/nccl_2.28.9-1+cuda13.0_x86_64.txz
## KV Cache Manager

The [KV cache manager](https://github.com/InternLM/lmdeploy/blob/main/src/turbomind/models/llama/SequenceManager.h) of TurboMind is a memory-pool-liked object that also implements LRU policy so that it can be viewed as a form of __cache of KV caches__. It works in the following way
The KV cache manager of TurboMind is a memory-pool-liked object that also implements LRU policy so that it can be viewed as a form of __cache of KV caches__. It works in the following way
@lvhan028
lvhan028 merged commit 67c18b9 into InternLM:main Jul 17, 2026
8 checks passed
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.

3 participants