# 获取方式1：
docker pull quay.io/ascend/sglang:v0.5.10-npu.rc1-910b

# 获取方式2：
docker pull m.daocloud.io/quay.io/ascend/sglang:v0.5.10-npu.rc1-910b

# 获取方式3：
docker pull quay.nju.edu.cn/ascend/sglang:v0.5.10-npu.rc1-910b

# 如果环境中没有安装modelscope，先执行以下命令
pip install modelscope

# --local_dir：按需修改为指定的存储路径，如/root/.cache/models/

modelscope download --model Eco-Tech/MiniMax-M2.5-w8a8-QuaRot --local_dir /root/.cache/models/MiniMax-M2.5-w8a8-QuaRot

# device 可按需挂载。示例为0-7卡 (Atlas A2: /dev/davinci[0-7] Atlas A3:/dev/davinci[0-15])
# 挂载目录需包含权重所在路径，示例为/root/.cache，可以根据实际情况指定
# 设置容器名称
export CONTAINER_NAME=MiniMax-M2.5-w8a8
# 选择镜像
export IMAGE=quay.io/ascend/sglang:v0.5.10-npu.rc1-910b

docker run --rm \
    --name $CONTAINER_NAME \
    --net=host \
    --shm-size=512g \
    --device /dev/davinci0 \
    --device /dev/davinci1 \
    --device /dev/davinci2 \
    --device /dev/davinci3 \
    --device /dev/davinci4 \
    --device /dev/davinci5 \
    --device /dev/davinci6 \
    --device /dev/davinci7 \
    --device /dev/davinci_manager \
    --device /dev/devmm_svm \
    --device /dev/hisi_hdc \
    -v /usr/local/dcmi:/usr/local/dcmi \
    -v /usr/local/Ascend/driver/tools/hccn_tool:/usr/local/Ascend/driver/tools/hccn_tool \
    -v /usr/local/bin/npu-smi:/usr/local/bin/npu-smi \
    -v /usr/local/Ascend/driver/lib64/:/usr/local/Ascend/driver/lib64/ \
    -v /usr/local/Ascend/driver/version.info:/usr/local/Ascend/driver/version.info \
    -v /etc/ascend_install.info:/etc/ascend_install.info \
    -v /etc/hccn.conf:/etc/hccn.conf \
    -v /root/.cache:/root/.cache \
    -it $IMAGE bash

sysctl -w vm.swappiness=0
sysctl -w kernel.numa_balancing=0
sysctl -w kernel.sched_migration_cost_ns=50000
# bind cpu
export SGLANG_SET_CPU_AFFINITY=1

unset https_proxy
unset http_proxy
unset HTTPS_PROXY
unset HTTP_PROXY
unset ASCEND_LAUNCH_BLOCKING
# cann
source /usr/local/Ascend/ascend-toolkit/set_env.sh
source /usr/local/Ascend/nnal/atb/set_env.sh

export STREAMS_PER_DEVICE=32
export HCCL_BUFFSIZE=1024
export HCCL_OP_EXPANSION_MODE=AIV


MODEL_PATH="/root/.cache/models/MiniMax-M2.5-w8a8-QuaRot"

sglang serve \
        --model-path $MODEL_PATH \
        --attention-backend ascend \
        --device npu \
        --tp-size 8 \
        --chunked-prefill-size 4096 --max-prefill-tokens 280000 \
        --disable-radix-cache \
        --trust-remote-code \
        --host 0.0.0.0 \
        --mem-fraction-static 0.9 \
        --port 8000 \
        --served-model-name mm2.5 \
        --tool-call-parser minimax-m2 \
        --reasoning-parser minimax-append-think \
        --cuda-graph-bs 16 \
        --quantization modelslim \
        --enable-multimodal \
        --mm-attention-backend ascend_attn \
        > log-mm2.5-serve.log 2>&1 &

curl http://localhost:8000/v1/chat/completions \
    -H "Content-Type: application/json" \
    -d '{
        "model": "mm2.5",
        "messages": [
            {"role": "system", "content": [{"type": "text", "text": "You are a helpful assistant."}]},
            {"role": "user", "content": [{"type": "text", "text": "Who won the world series in 2020?"}]}
        ]
    }'

{"id":"4e543ea1e70947d79722a9a531d1daf8","object":"chat.completion","created":1778118242,"model":"mm2.5","choices":[{"index":0,"message":{"role":"assistant","content":"<think>The user asks: \"Who won the world series in 2020?\" That's a straightforward question: The Los Angeles Dodgers won the 2020 World Series. I should respond with that answer. No need for any extra content. There's no policy violation. Just answer.\n</think>\n\nThe Los Angeles Dodgers won the 2020 World Series.","reasoning_content":null,"tool_calls":null},"logprobs":null,"finish_reason":"stop","matched_stop":200020}],"usage":{"prompt_tokens":32,"total_tokens":106,"completion_tokens":74,"prompt_tokens_details":null,"reasoning_tokens":55},"metadata":{"weight_version":"default"}}