💫 MetaX（沐曦股份）metax3.6

Based on Triton 3.6, x64
Available for C550

1. Quick start

1.1 Use the image (C550)

Tips: You can use apt install aria2; aria2c -c ... instead of wget ... to download large files with resume support.

IMAGE=harbor.baai.ac.cn/flagtree/flagtree-metax-py312-torch2.8.0-metax3.7.0.7-ubuntu22.04:20261001-3.6-vllm0.24.0
# Plan A: docker pull (45.2GB)
docker pull ${IMAGE}
# Plan B: docker load (18GB)
wget https://baai-cp-web.ks3-cn-beijing.ksyuncs.com/trans/flagtree-metax-py312-torch2.8.0-metax3.7.0.7-ubuntu22.04.20261001-3.6-vllm0.24.0.tar.gz
docker load -i flagtree-metax-py312-torch2.8.0-metax3.7.0.7-ubuntu22.04.20261001-3.6-vllm0.24.0.tar.gz

CONTAINER=flagtree-dev-xxx
docker run -dit \
    --net=host --uts=host --ipc=host --privileged=true \
    --group-add video \
    --shm-size 100gb --ulimit memlock=-1 \
    --security-opt seccomp=unconfined --security-opt apparmor=unconfined \
    --device=/dev/dri --device=/dev/mxcd \
    -v /etc/localtime:/etc/localtime:ro \
    -v /data:/data -v /home:/home \
    -w /root --name ${CONTAINER} ${IMAGE} bash
docker exec -it ${CONTAINER} /bin/bash

1.2 Source-free Installation

# Note: First install PyTorch, then execute the following commands
python3 -m pip uninstall -y triton  # Repeat the cmd until fully uninstalled
RES="--index-url=https://resource.flagos.net/repository/flagos-pypi-hosted/simple"
python3.12 -m pip install flagtree===0.7.0.post2+metax3.6 $RES

2. Build from Source

2.1 Manually download the FlagTree dependencies

If your network connection is available, you do not need to download the dependencies which will be fetched automatically during the build.

mkdir -p ~/.flagtree/metax; cd ~/.flagtree/metax
wget https://baai-cp-web.ks3-cn-beijing.ksyuncs.com/trans/metax-llvm19-3.8.0.6-x86_64_v0.6.0.tar.gz
tar zxvf metax-llvm19-3.8.0.6-x86_64_v0.6.0.tar.gz
wget https://baai-cp-web.ks3-cn-beijing.ksyuncs.com/trans/metaxTritonPlugin-cpython3.12-x86_64_v0.6.1.tar.gz
tar zxvf metaxTritonPlugin-cpython3.12-x86_64_v0.6.1.tar.gz

2.2 Manually download the Triton dependencies

The Triton dependencies are already downloaded and installed in the image. If your network connection is available, you do not need to download the dependencies which will be fetched automatically during the build.

cd ${YOUR_CODE_DIR}/FlagTree
# For Triton 3.6 (x64)
wget https://baai-cp-web.ks3-cn-beijing.ksyuncs.com/trans/build-deps-triton_3.6.x-linux-x64.tar.gz
sh python/scripts/unpack_triton_build_deps.sh ./build-deps-triton_3.6.x-linux-x64.tar.gz

After executing the above script, the original ~/.triton directory will be renamed, and a new ~/.triton directory will be created to store the pre-downloaded packages. Note that the script will prompt for manual confirmation during execution.

2.3 Build Commands

cd ${YOUR_CODE_DIR}/FlagTree
git checkout main
export FLAGTREE_BACKEND=metax
MAX_JOBS=32 python3 -m pip install . --no-build-isolation -v

3. Testing and validation

After installing flagtree, you can check it with:

python3 -m pip show flagtree

Refer to Tests of metax3.6 backend

4. Run Qwen vLLM benchmark

4.1 Install FlagOS

Install flagtree as described above.

Install flag_gems and vllm-plugin-fl as follows.

git clone https://github.com/flagos-ai/FlagGems.git
cd FlagGems; git checkout v5.3.4
python3 -m pip install --no-build-isolation .

git clone https://github.com/flagos-ai/vllm-plugin-FL.git
cd vllm-plugin-FL; git checkout 0.3.0rc2.post3
python3 -m pip install --no-build-isolation .

4.2 Download the model

python3 -m modelscope.cli.cli download --model Qwen/Qwen3.6-27B \
    --local_dir ${MODEL_DIR}/Qwen3.6-27B
python3 -m modelscope.cli.cli download --model Qwen/Qwen3.6-35B-A3B \
    --local_dir ${MODEL_DIR}/Qwen3.6-35B-A3B

4.3 Run the benchmark

# ~/env.sh
export CUDA_VISIBLE_DEVICES=0,1  # C550 * 2 for Qwen3.6-27B
export VLLM_QWEN3_PORT=7000

export BENCH_SCRIPT_DIR="${GIT_DIR}/FlagTree/.github/workflows/benchmark/metax/metax3.6-qwen3.6"
cd ${MODEL_DIR}
vim ${BENCH_SCRIPT_DIR}/start.sh  # Edit vllm serve Qwen3.6-27B or Qwen3.6-35B-A3B
bash ${BENCH_SCRIPT_DIR}/start.sh
bash ${BENCH_SCRIPT_DIR}/test.sh
vim ${BENCH_SCRIPT_DIR}/all_perf.py  # Edit TOKENIZER_PATH
bash ${BENCH_SCRIPT_DIR}/perf.sh
bash ${BENCH_SCRIPT_DIR}/stop.sh

Refer to Qwen Benchmark of metax3.6 backend

💫 MetaX（沐曦股份）metax3.0

Based on Triton 3.0, x64
Available for C550

1. Quick start

1.1 Use the preinstalled image (C550)

If you use this preinstalled image, you do not need to perform the later installation steps.

# Plan A: docker pull (28.1GB)
IMAGE=harbor.baai.ac.cn/flagtree/flagtree-metax-py312-torch2.8.0-vllm0.15.0-metax3.5.3.x-ubuntu22.04:202604-0.5.1
docker pull ${IMAGE}
# Plan B: docker load (8.1GB)
IMAGE=flagtree-metax-py312-torch2.8.0-vllm0.15.0-metax3.5.3.x-ubuntu22.04:202604-0.5.1
wget https://baai-cp-web.ks3-cn-beijing.ksyuncs.com/trans/flagtree-metax-py312-torch2.8.0-vllm0.15.0-metax3.5.3.x-ubuntu22.04.202604-0.5.1.tar.gz
docker load -i flagtree-metax-py312-torch2.8.0-vllm0.15.0-metax3.5.3.x-ubuntu22.04.202604-0.5.1.tar.gz

CONTAINER=flagtree-dev-xxx
docker run -dit \
    --net=host --uts=host --ipc=host --privileged=true \
    --group-add video \
    --shm-size 100gb --ulimit memlock=-1 \
    --security-opt seccomp=unconfined --security-opt apparmor=unconfined \
    --device=/dev/dri --device=/dev/mxcd \
    -v /etc/localtime:/etc/localtime:ro \
    -v /data:/data -v /home:/home \
    -w /root --name ${CONTAINER} ${IMAGE} bash
docker exec -it ${CONTAINER} /bin/bash

1.2 Source-free Installation

# Note: First install PyTorch, then execute the following commands
python3 -m pip uninstall -y triton  # Repeat the cmd until fully uninstalled
RES="--index-url=https://resource.flagos.net/repository/flagos-pypi-hosted/simple"
python3.12 -m pip install flagtree===0.5.1+metax3.0 $RES

flagtree is already installed in the preinstalled image.

2. Build from Source

2.1 Manually download the FlagTree dependencies

If your network connection is available, you do not need to download the dependencies which will be fetched automatically during the build.

mkdir -p ~/.flagtree/metax; cd ~/.flagtree/metax
wget https://baai-cp-web.ks3-cn-beijing.ksyuncs.com/trans/metaxTritonPlugin-cpython3.12-glibc2.35-glibcxx3.4.30-cxxabi1.3.13-linux-x86_64_v0.5.0.tar.gz
tar zxvf metaxTritonPlugin-cpython3.12-glibc2.35-glibcxx3.4.30-cxxabi1.3.13-linux-x86_64_v0.5.0.tar.gz
wget https://baai-cp-web.ks3-cn-beijing.ksyuncs.com/trans/ext_maca_mathlib_bc_v0.5.0.tar.gz
tar zxvf ext_maca_mathlib_bc_v0.5.0.tar.gz
# NOTE: Contact the Metax vendor to obtain maca-llvm-metax20250708.521-x86_64.tar.xz
tar xvf maca-llvm-metax20250708.521-x86_64.tar.xz

2.2 Manually download the Triton dependencies

The Triton dependencies are already downloaded and installed in the preinstalled image. If your network connection is available, you do not need to download the dependencies which will be fetched automatically during the build.

cd ${YOUR_CODE_DIR}/FlagTree
# For Triton 3.1 (x64)
wget https://baai-cp-web.ks3-cn-beijing.ksyuncs.com/trans/build-deps-triton_3.1.x-linux-x64.tar.gz
sh python/scripts/unpack_triton_build_deps.sh ./build-deps-triton_3.1.x-linux-x64.tar.gz

2.3 Build Commands

cd ${YOUR_CODE_DIR}/FlagTree/python
git checkout -b triton_v3.1.x origin/triton_v3.1.x
export FLAGTREE_BACKEND=metax
MAX_JOBS=32 python3 -m pip install . --no-build-isolation -v

3. Testing and validation

You can check the installed flagtree with:

python3 -m pip show flagtree

Refer to Tests of metax3.0 backend

User manual for metax - flagos-ai/FlagTree GitHub Wiki

💫 MetaX（沐曦股份）metax3.6

1. Quick start

2. Build from Source

3. Testing and validation

4. Run Qwen vLLM benchmark

💫 MetaX（沐曦股份）metax3.0

1. Quick start

2. Build from Source

3. Testing and validation

⚠️ GitHub.com Fallback ⚠️

User manual for metax - flagos-ai/FlagTree GitHub Wiki

💫 MetaX（沐曦股份）metax3.6

1. Quick start

2. Build from Source

3. Testing and validation

4. Run Qwen vLLM benchmark

💫 MetaX（沐曦股份）metax3.0

1. Quick start

2. Build from Source

3. Testing and validation

⚠️ **GitHub.com Fallback** ⚠️

⚠️ GitHub.com Fallback ⚠️