DeepSeek-V3.2

1 Introduction

DeepSeek-V3.2 是一种稀疏注意力模型。其主要架构与 DeepSeek-V3.1 类似，但引入了稀疏注意力机制，旨在探索和验证长上下文场景下训练与推理效率的优化方案。

The DeepSeek-V3.2 model is first supported in vllm-ascend:v0.13.0rc1, and all v0.13.0rc1 and later versions can run stably. To use the latest features (e.g., PD separation, MTP), it is recommended to use the latest release candidate or official version.

本文档将展示该模型的主要验证步骤，包括支持的特性、特性配置、环境准备、单节点与多节点部署、精度及性能评估。

2 Supported Features

请参考支持的特性获取该模型支持的特性矩阵。

请参考特性指南获取特性的配置方法。

3 Prerequisites

3.1 Model Weight

• DeepSeek-V3.2-Exp-W8A8（量化版本）：需要 1 个 Atlas 800 A3（64G × 16）节点 或 2 个 Atlas 800 A2（64G × 8）节点。下载模型权重
• DeepSeek-V3.2-w8a8（量化版本）：需要 1 个 Atlas 800 A3（64G × 16）节点 或 2 个 Atlas 800 A2（64G × 8）节点。下载模型权重

建议将模型权重下载到多节点的共享目录中，例如 /root/.cache/。

3.2 Verify Multi-node Communication (Optional)

如需部署多节点环境，需按照验证多节点通信环境中的说明验证多节点通信。

4 Installation

4.1 Docker Image Installation

您可以直接使用官方 Docker 镜像运行 DeepSeek-V3.2。

A3 系列A2 系列

在每个节点上启动 Docker 镜像。

export IMAGE=quay.io/ascend/vllm-ascend:|vllm_ascend_version|-a3
docker run --rm \
    --name vllm-ascend \
    --shm-size=1g \
    --net=host \
    --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/davinci8 \
    --device /dev/davinci9 \
    --device /dev/davinci10 \
    --device /dev/davinci11 \
    --device /dev/davinci12 \
    --device /dev/davinci13 \
    --device /dev/davinci14 \
    --device /dev/davinci15 \
    --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 /root/.cache:/root/.cache \
    -it $IMAGE bash

在每个节点上启动 Docker 镜像。

export IMAGE=quay.io/ascend/vllm-ascend:|vllm_ascend_version|
docker run --rm \
    --name vllm-ascend \
    --shm-size=1g \
    --net=host \
    --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 /root/.cache:/root/.cache \
    -it $IMAGE bash

To verify the successful installation of the environment, please refer to installation.

如需部署多节点环境，需要在每个节点上进行环境设置。

4.2 Source Code Installation

此外，如果您不想使用上述 Docker 镜像，也可以从源码构建所有组件：

• 从源码安装 vllm-ascend，请参考安装指南。

5 Online Service Deployment

Notice: 在本教程中，我们假设您已将模型权重下载到 /root/.cache/。您可以根据需要更改为自己的路径。

5.1 Single-node Deployment

• 量化模型 DeepSeek-V3.2-w8a8 可部署在 1 个 Atlas 800 A3（64G × 16）上。

运行以下脚本执行在线推理。

export HCCL_OP_EXPANSION_MODE="AIV"
export OMP_PROC_BIND=false
export OMP_NUM_THREADS=10
export VLLM_USE_V1=1
export HCCL_BUFFSIZE=200
export VLLM_ASCEND_ENABLE_MLAPO=1
export PYTORCH_NPU_ALLOC_CONF=expandable_segments:True
export VLLM_ASCEND_ENABLE_FLASHCOMM1=1

vllm serve /root/.cache/modelscope/hub/models/vllm-ascend/DeepSeek-V3.2-W8A8 \
--host 0.0.0.0 \
--port 8000 \
--data-parallel-size 2 \
--tensor-parallel-size 8 \
--quantization ascend \
--seed 1024 \
--served-model-name deepseek_v3_2 \
--enable-expert-parallel \
--max-num-seqs 16 \
--max-model-len 8192 \
--max-num-batched-tokens 4096 \
--trust-remote-code \
--no-enable-prefix-caching \
--gpu-memory-utilization 0.92 \
--compilation-config '{"cudagraph_mode": "FULL_DECODE_ONLY"}' \
--speculative-config '{"num_speculative_tokens": 3, "method": "deepseek_mtp"}'

Common Issues Tip: If you encounter issues, Refer to FAQs.

5.2 Multi-Node Data Parallel Deployment

• DeepSeek-V3.2-w8a8：至少需要 2 个 Atlas 800 A2（64G × 8）。

Common Issues Tip: If you encounter issues, Refer to FAQs.

分别在两个节点上运行以下脚本。

A3 系列A2 系列

节点0

# this obtained through ifconfig
# nic_name is the network interface name corresponding to local_ip of the current node
nic_name="xxx"
local_ip="xxx"

# The value of node0_ip must be consistent with the value of local_ip set in node0 (master node)
node0_ip="xxxx"

export HCCL_OP_EXPANSION_MODE="AIV"

export HCCL_IF_IP=$local_ip
export GLOO_SOCKET_IFNAME=$nic_name
export TP_SOCKET_IFNAME=$nic_name
export HCCL_SOCKET_IFNAME=$nic_name
export OMP_PROC_BIND=false
export OMP_NUM_THREADS=10
export VLLM_USE_V1=1
export HCCL_BUFFSIZE=200
export VLLM_ASCEND_ENABLE_MLAPO=1
export PYTORCH_NPU_ALLOC_CONF=expandable_segments:True
export VLLM_ASCEND_ENABLE_FLASHCOMM1=1

vllm serve /root/.cache/modelscope/hub/models/vllm-ascend/DeepSeek-V3.2-W8A8 \
--host 0.0.0.0 \
--port 8077 \
--data-parallel-size 2 \
--data-parallel-size-local 1 \
--data-parallel-address $node0_ip \
--data-parallel-rpc-port 12890 \
--tensor-parallel-size 16 \
--quantization ascend \
--seed 1024 \
--served-model-name deepseek_v3_2 \
--enable-expert-parallel \
--max-num-seqs 16 \
--max-model-len 8192 \
--max-num-batched-tokens 4096 \
--trust-remote-code \
--no-enable-prefix-caching \
--gpu-memory-utilization 0.92 \
--compilation-config '{"cudagraph_mode": "FULL_DECODE_ONLY"}' \
--speculative-config '{"num_speculative_tokens": 3, "method": "deepseek_mtp"}'

节点1

# this obtained through ifconfig
# nic_name is the network interface name corresponding to local_ip of the current node
nic_name="xxx"
local_ip="xxx"

# The value of node0_ip must be consistent with the value of local_ip set in node0 (master node)
node0_ip="xxxx"

export HCCL_OP_EXPANSION_MODE="AIV"

export HCCL_IF_IP=$local_ip
export GLOO_SOCKET_IFNAME=$nic_name
export TP_SOCKET_IFNAME=$nic_name
export HCCL_SOCKET_IFNAME=$nic_name
export OMP_PROC_BIND=false
export OMP_NUM_THREADS=10
export VLLM_USE_V1=1
export HCCL_BUFFSIZE=200
export VLLM_ASCEND_ENABLE_MLAPO=1
export PYTORCH_NPU_ALLOC_CONF=expandable_segments:True
export VLLM_ASCEND_ENABLE_FLASHCOMM1=1

vllm serve /root/.cache/modelscope/hub/models/vllm-ascend/DeepSeek-V3.2-W8A8 \
--host 0.0.0.0 \
--port 8077 \
--headless \
--data-parallel-size 2 \
--data-parallel-size-local 1 \
--data-parallel-start-rank 1 \
--data-parallel-address $node0_ip \
--data-parallel-rpc-port 12890 \
--tensor-parallel-size 16 \
--quantization ascend \
--seed 1024 \
--served-model-name deepseek_v3_2 \
--enable-expert-parallel \
--max-num-seqs 16 \
--max-model-len 8192 \
--max-num-batched-tokens 4096 \
--trust-remote-code \
--no-enable-prefix-caching \
--gpu-memory-utilization 0.92 \
--compilation-config '{"cudagraph_mode": "FULL_DECODE_ONLY"}' \
--speculative-config '{"num_speculative_tokens": 3, "method": "deepseek_mtp"}'

节点0

# this obtained through ifconfig
# nic_name is the network interface name corresponding to local_ip of the current node
nic_name="xxx"
local_ip="xxx"

# The value of node0_ip must be consistent with the value of local_ip set in node0 (master node)
node0_ip="xxxx"

export HCCL_OP_EXPANSION_MODE="AIV"

export HCCL_IF_IP=$local_ip
export GLOO_SOCKET_IFNAME=$nic_name
export TP_SOCKET_IFNAME=$nic_name
export HCCL_SOCKET_IFNAME=$nic_name
export OMP_PROC_BIND=false
export OMP_NUM_THREADS=100
export VLLM_USE_V1=1
export HCCL_BUFFSIZE=200
export VLLM_ASCEND_ENABLE_MLAPO=1
export PYTORCH_NPU_ALLOC_CONF=expandable_segments:True
export VLLM_ASCEND_ENABLE_FLASHCOMM1=1
export HCCL_CONNECT_TIMEOUT=120
export HCCL_INTRA_PCIE_ENABLE=1
export HCCL_INTRA_ROCE_ENABLE=0

vllm serve /root/.cache/modelscope/hub/models/vllm-ascend/DeepSeek-V3.2-W8A8 \
--host 0.0.0.0 \
--port 8077 \
--data-parallel-size 2 \
--data-parallel-size-local 1 \
--data-parallel-address $node0_ip \
--data-parallel-rpc-port 13389 \
--tensor-parallel-size 8 \
--quantization ascend \
--seed 1024 \
--served-model-name deepseek_v3_2 \
--enable-expert-parallel \
--max-num-seqs 16 \
--max-model-len 8192 \
--max-num-batched-tokens 4096 \
--trust-remote-code \
--no-enable-prefix-caching \
--gpu-memory-utilization 0.92 \
--compilation-config '{"cudagraph_mode": "FULL_DECODE_ONLY", "cudagraph_capture_sizes":[8, 16, 24, 32, 40, 48]}' \
--speculative-config '{"num_speculative_tokens": 3, "method": "deepseek_mtp"}'

节点1

# this obtained through ifconfig
# nic_name is the network interface name corresponding to local_ip of the current node
nic_name="xxx"
local_ip="xxx"

# The value of node0_ip must be consistent with the value of local_ip set in node0 (master node)
node0_ip="xxxx"

export HCCL_OP_EXPANSION_MODE="AIV"

export HCCL_IF_IP=$local_ip
export GLOO_SOCKET_IFNAME=$nic_name
export TP_SOCKET_IFNAME=$nic_name
export HCCL_SOCKET_IFNAME=$nic_name
export OMP_PROC_BIND=false
export OMP_NUM_THREADS=100
export VLLM_USE_V1=1
export HCCL_BUFFSIZE=200
export VLLM_ASCEND_ENABLE_MLAPO=1
export PYTORCH_NPU_ALLOC_CONF=expandable_segments:True
export VLLM_ASCEND_ENABLE_FLASHCOMM1=1
export HCCL_CONNECT_TIMEOUT=120
export HCCL_INTRA_PCIE_ENABLE=1
export HCCL_INTRA_ROCE_ENABLE=0

vllm serve /root/.cache/modelscope/hub/models/vllm-ascend/DeepSeek-V3.2-W8A8 \
--host 0.0.0.0 \
--port 8077 \
--headless \
--data-parallel-size 2 \
--data-parallel-size-local 1 \
--data-parallel-start-rank 1 \
--data-parallel-address $node0_ip \
--data-parallel-rpc-port 13389 \
--tensor-parallel-size 8 \
--quantization ascend \
--seed 1024 \
--served-model-name deepseek_v3_2 \
--enable-expert-parallel \
--max-num-seqs 16 \
--max-model-len 8192 \
--max-num-batched-tokens 4096 \
--trust-remote-code \
--no-enable-prefix-caching \
--gpu-memory-utilization 0.92 \
--compilation-config '{"cudagraph_mode": "FULL_DECODE_ONLY", "cudagraph_capture_sizes":[8, 16, 24, 32, 40, 48]}' \
--speculative-config '{"num_speculative_tokens": 3, "method": "deepseek_mtp"}'

Notice: The parameters are explained as follows:

• For multi-node deployment, we recommend using dp2tp16 (A3) or dp2tp8 (A2) with expert parallel enabled.

5.3 Multi-Node PD Separation Deployment

We recommend using Mooncake for deployment: Mooncake.

In the standard single-node deployment mode, Prefill (prompt processing) and Decode (token generation) tasks run on the same set of NPUs. PD (Prefill-Decode) separation addresses this by running Prefill and Decode on dedicated node groups, each configured independently:

• Prefill nodes focus on high-throughput prompt processing, optimized for compute and communication.
• Decode nodes focus on low-latency token generation, optimized for memory bandwidth.

This architecture is recommended for production deployments with concurrent multi-user workloads, where stable latency and high throughput are both required.

我们将展示 DeepSeek-V3.2 在多节点环境下采用 1P1D 配置以获得更好性能的部署指南。

Common Issues Tip: If you encounter issues, Refer to FAQs.

• DeepSeek-V3.2-w8a8-mtp-QuaRot 1P1D Layerwise require 4 Atlas 800 A3 (64G × 16).

To run the vllm-ascend Prefill-Decode Disaggregation service, you need to deploy a launch_online_dp.py script and a run_dp_template.sh script on each node and deploy a proxy.sh script on prefill master node to forward requests.

launch_online_dp.py

Parameter descriptions:

Parameter	Type	Required	Default	Description
--dp-size	int	Yes	-	Data parallel size (total number of DP ranks across all nodes).
--tp-size	int	No	1	Tensor parallel size within each DP rank.
--dp-size-local	int	No	(same as --dp-size)	Number of DP ranks on the current node. If not set, defaults to --dp-size.
--dp-rank-start	int	No	0	Starting rank offset for data parallel ranks on this node.
--dp-address	str	Yes	-	IP address of the data parallel master node (node 0).
--dp-rpc-port	str	No	12345	RPC port for data parallel master communication.
--vllm-start-port	int	No	9000	Starting port for each vLLM engine instance on this node. Each DP rank's engine port = vllm_start_port + local rank index.

1. run_dp_template.sh script

Node 0(Prefill)Node 1(Prefill)Node 0(Decode)Node 1(Decode)

nic_name="enp48s3u1u1" # change to your own nic name
local_ip=141.61.39.105 # change to your own ip

export HCCL_OP_EXPANSION_MODE="AIV"

export HCCL_IF_IP=$local_ip
export GLOO_SOCKET_IFNAME=$nic_name
export TP_SOCKET_IFNAME=$nic_name
export HCCL_SOCKET_IFNAME=$nic_name

export OMP_PROC_BIND=false
export OMP_NUM_THREADS=10
export PYTORCH_NPU_ALLOC_CONF=expandable_segments:True
export VLLM_USE_V1=1
export HCCL_BUFFSIZE=256

export ASCEND_AGGREGATE_ENABLE=1
export ASCEND_TRANSPORT_PRINT=1
export ACL_OP_INIT_MODE=1
export ASCEND_A3_ENABLE=1
# Timeout (in seconds) for automatically releasing the prefiller’s KV cache for a particular request.
export VLLM_MOONCAKE_ABORT_REQUEST_TIMEOUT=480

export ASCEND_RT_VISIBLE_DEVICES=$1

export VLLM_ASCEND_ENABLE_FLASHCOMM1=1


vllm serve /root/.cache/Eco-Tech/DeepSeek-V3.2-w8a8-mtp-QuaRot \
    --host 0.0.0.0 \
    --port $2 \
    --data-parallel-size $3 \
    --data-parallel-rank $4 \
    --data-parallel-address $5 \
    --data-parallel-rpc-port $6 \
    --tensor-parallel-size $7 \
    --enable-expert-parallel \
    --speculative-config '{"num_speculative_tokens": 2, "method":"deepseek_mtp"}' \
    --profiler-config \
    '{"profiler": "torch",
    "torch_profiler_dir": "./vllm_profile",
    "torch_profiler_with_stack": false}' \
    --seed 1024 \
    --served-model-name deepseek_v3.2 \
    --max-model-len 68000 \
    --max-num-batched-tokens 32560 \
    --trust-remote-code \
    --max-num-seqs 64 \
    --gpu-memory-utilization 0.82 \
    --quantization ascend \
    --enforce-eager \
    --no-enable-prefix-caching \
    --additional-config '{"layer_sharding": ["q_b_proj", "o_proj"], "enable_dsa_cp": true}' \
    --kv-transfer-config \
    '{"kv_connector": "MooncakeConnectorV1",
    "kv_role": "kv_producer",
    "kv_port": "30000",
    "kv_connector_extra_config": {
                "prefill": {
                        "dp_size": 2,
                        "tp_size": 16
                },
                "decode": {
                        "dp_size": 8,
                        "tp_size": 4
                }
        }
    }'

nic_name="enp48s3u1u1" # change to your own nic name
local_ip=141.61.39.113 # change to your own ip

export HCCL_OP_EXPANSION_MODE="AIV"

export HCCL_IF_IP=$local_ip
export GLOO_SOCKET_IFNAME=$nic_name
export TP_SOCKET_IFNAME=$nic_name
export HCCL_SOCKET_IFNAME=$nic_name

export OMP_PROC_BIND=false
export OMP_NUM_THREADS=10
export PYTORCH_NPU_ALLOC_CONF=expandable_segments:True
export VLLM_USE_V1=1
export HCCL_BUFFSIZE=256

export ASCEND_AGGREGATE_ENABLE=1
export ASCEND_TRANSPORT_PRINT=1
export ACL_OP_INIT_MODE=1
export ASCEND_A3_ENABLE=1
# Timeout (in seconds) for automatically releasing the prefiller’s KV cache for a particular request.
export VLLM_MOONCAKE_ABORT_REQUEST_TIMEOUT=480

export ASCEND_RT_VISIBLE_DEVICES=$1

export VLLM_ASCEND_ENABLE_FLASHCOMM1=1


vllm serve /root/.cache/Eco-Tech/DeepSeek-V3.2-w8a8-mtp-QuaRot \
    --host 0.0.0.0 \
    --port $2 \
    --data-parallel-size $3 \
    --data-parallel-rank $4 \
    --data-parallel-address $5 \
    --data-parallel-rpc-port $6 \
    --tensor-parallel-size $7 \
    --enable-expert-parallel \
    --speculative-config '{"num_speculative_tokens": 2, "method":"deepseek_mtp"}' \
    --profiler-config \
    '{"profiler": "torch",
    "torch_profiler_dir": "./vllm_profile",
    "torch_profiler_with_stack": false}' \
    --seed 1024 \
    --served-model-name deepseek_v3.2 \
    --max-model-len 68000 \
    --max-num-batched-tokens 32560 \
    --trust-remote-code \
    --max-num-seqs 64 \
    --gpu-memory-utilization 0.82 \
    --quantization ascend \
    --enforce-eager \
    --no-enable-prefix-caching \
    --additional-config '{"layer_sharding": ["q_b_proj", "o_proj"], "enable_dsa_cp": true}' \
    --kv-transfer-config \
    '{"kv_connector": "MooncakeConnectorV1",
    "kv_role": "kv_producer",
    "kv_port": "30000",
    "kv_connector_extra_config": {
                "prefill": {
                        "dp_size": 2,
                        "tp_size": 16
                },
                "decode": {
                        "dp_size": 8,
                        "tp_size": 4
                }
        }
    }'

nic_name="enp48s3u1u1" # change to your own nic name
local_ip=141.61.39.117 # change to your own ip

export HCCL_OP_EXPANSION_MODE="AIV"

export HCCL_IF_IP=$local_ip
export GLOO_SOCKET_IFNAME=$nic_name
export TP_SOCKET_IFNAME=$nic_name
export HCCL_SOCKET_IFNAME=$nic_name

#Mooncake
export OMP_PROC_BIND=false
export OMP_NUM_THREADS=10

export PYTORCH_NPU_ALLOC_CONF=expandable_segments:True
export VLLM_USE_V1=1
export HCCL_BUFFSIZE=256


export ASCEND_AGGREGATE_ENABLE=1
export ASCEND_TRANSPORT_PRINT=1
export ACL_OP_INIT_MODE=1
export ASCEND_A3_ENABLE=1
# Timeout (in seconds) for automatically releasing the prefiller’s KV cache for a particular request.
export VLLM_MOONCAKE_ABORT_REQUEST_TIMEOUT=480

export TASK_QUEUE_ENABLE=1

export ASCEND_RT_VISIBLE_DEVICES=$1


vllm serve /root/.cache/Eco-Tech/DeepSeek-V3.2-w8a8-mtp-QuaRot \
    --host 0.0.0.0 \
    --port $2 \
    --data-parallel-size $3 \
    --data-parallel-rank $4 \
    --data-parallel-address $5 \
    --data-parallel-rpc-port $6 \
    --tensor-parallel-size $7 \
    --enable-expert-parallel \
    --speculative-config '{"num_speculative_tokens": 2, "method":"deepseek_mtp"}' \
    --profiler-config \
    '{"profiler": "torch",
    "torch_profiler_dir": "./vllm_profile",
    "torch_profiler_with_stack": false}' \
    --seed 1024 \
    --served-model-name deepseek_v3.2 \
    --max-model-len 68000 \
    --max-num-batched-tokens 12 \
    --compilation-config '{"cudagraph_mode":"FULL_DECODE_ONLY", "cudagraph_capture_sizes":[3, 6, 9, 12]}' \
    --trust-remote-code \
    --max-num-seqs 4 \
    --gpu-memory-utilization 0.95 \
    --no-enable-prefix-caching \
    --quantization ascend \
    --kv-transfer-config \
    '{"kv_connector": "MooncakeConnectorV1",
    "kv_role": "kv_consumer",
    "kv_port": "30100",
    "kv_connector_extra_config": {
                "prefill": {
                        "dp_size": 2,
                        "tp_size": 16
                },
                "decode": {
                        "dp_size": 8,
                        "tp_size": 4
                }
        }
    }' \
    --additional-config '{"recompute_scheduler_enable" : true}'

nic_name="enp48s3u1u1" # change to your own nic name
local_ip=141.61.39.181 # change to your own ip

export HCCL_OP_EXPANSION_MODE="AIV"

export HCCL_IF_IP=$local_ip
export GLOO_SOCKET_IFNAME=$nic_name
export TP_SOCKET_IFNAME=$nic_name
export HCCL_SOCKET_IFNAME=$nic_name

#Mooncake
export OMP_PROC_BIND=false
export OMP_NUM_THREADS=10

export PYTORCH_NPU_ALLOC_CONF=expandable_segments:True
export VLLM_USE_V1=1
export HCCL_BUFFSIZE=256

export ASCEND_AGGREGATE_ENABLE=1
export ASCEND_TRANSPORT_PRINT=1
export ACL_OP_INIT_MODE=1
export ASCEND_A3_ENABLE=1
# Timeout (in seconds) for automatically releasing the prefiller’s KV cache for a particular request.
export VLLM_MOONCAKE_ABORT_REQUEST_TIMEOUT=480

export TASK_QUEUE_ENABLE=1

export ASCEND_RT_VISIBLE_DEVICES=$1


vllm serve /root/.cache/Eco-Tech/DeepSeek-V3.2-w8a8-mtp-QuaRot \
    --host 0.0.0.0 \
    --port $2 \
    --data-parallel-size $3 \
    --data-parallel-rank $4 \
    --data-parallel-address $5 \
    --data-parallel-rpc-port $6 \
    --tensor-parallel-size $7 \
    --enable-expert-parallel \
    --speculative-config '{"num_speculative_tokens": 2, "method":"deepseek_mtp"}' \
    --profiler-config \
    '{"profiler": "torch",
    "torch_profiler_dir": "./vllm_profile",
    "torch_profiler_with_stack": false}' \
    --seed 1024 \
    --served-model-name deepseek_v3.2 \
    --max-model-len 68000 \
    --max-num-batched-tokens 12 \
    --compilation-config '{"cudagraph_mode":"FULL_DECODE_ONLY",  "cudagraph_capture_sizes":[3, 6, 9, 12]}' \
    --trust-remote-code \
    --max-num-seqs 4 \
    --gpu-memory-utilization 0.95 \
    --no-enable-prefix-caching \
    --quantization ascend \
    --kv-transfer-config \
    '{"kv_connector": "MooncakeConnectorV1",
    "kv_role": "kv_consumer",
    "kv_port": "30100",
    "kv_connector_extra_config": {
                "prefill": {
                        "dp_size": 2,
                        "tp_size": 16
                },
                "decode": {
                        "dp_size": 8,
                        "tp_size": 4
                }
        }
    }' \
    --additional-config '{"recompute_scheduler_enable" : true}'

准备工作完成后，您可以在每个节点上使用以下命令启动服务器： 请参考大规模专家并行分布式 DP 服务器获取详细的启动方法。

1. 预填充节点 0

   # change ip to your own
   python launch_online_dp.py --dp-size 2 --tp-size 16 --dp-size-local 1 --dp-rank-start 0 --dp-address 141.61.39.105 --dp-rpc-port 12890 --vllm-start-port 9100
   

2. 预填充节点 1

   # change ip to your own
   python launch_online_dp.py --dp-size 2 --tp-size 16 --dp-size-local 1 --dp-rank-start 1 --dp-address 141.61.39.105 --dp-rpc-port 12890 --vllm-start-port 9100
   

3. 解码节点 0

   # change ip to your own
   python launch_online_dp.py --dp-size 8 --tp-size 4 --dp-size-local 4 --dp-rank-start 0 --dp-address 141.61.39.117 --dp-rpc-port 12777 --vllm-start-port 9100
   

4. 解码节点 1

   # change ip to your own
   python launch_online_dp.py --dp-size 8 --tp-size 4 --dp-size-local 4 --dp-rank-start 4 --dp-address 141.61.39.117 --dp-rpc-port 12777 --vllm-start-port 9100
   

Notice: To support a long context window on the stage of prefill, the parameter "layer_sharding": ["q_b_proj", "o_proj"] and "enable_dsa_cp": true needs to be added to --additional-config on each prefill node.

请求转发

要设置请求转发，请在任意机器上运行以下脚本。您可以在仓库的示例中找到代理程序：load_balance_proxy_layerwise_server_example.py

unset http_proxy
unset https_proxy

python load_balance_proxy_layerwise_server_example.py \
    --port 8000 \
    --host 141.61.39.105 \
    --prefiller-hosts \
       141.61.39.105 \
       141.61.39.113 \
    --prefiller-ports \
       9100 \
       9100 \
    --decoder-hosts \
      141.61.39.117 \
      141.61.39.117 \
      141.61.39.117 \
      141.61.39.117 \
      141.61.39.181 \
      141.61.39.181 \
      141.61.39.181 \
      141.61.39.181 \
    --decoder-ports \
      9100 9101 9102 9103 \
      9100 9101 9102 9103 \

6 Functional Verification

服务器启动后，您可以使用输入提示词查询模型：

注意：

• <node0_ip>：运行服务器的节点 IP 地址（例如 localhost）。对于 PD 分离部署，请使用代理脚本所在节点的主机 IP。
• <port>：服务器启动命令中指定的端口号（例如 8000）。对于 PD 分离部署，请使用代理脚本中配置的端口。

curl http://<node0_ip>:<port>/v1/completions \
    -H "Content-Type: application/json" \
    -d '{
        "model": "deepseek_v3.2",
        "prompt": "The future of AI is",
        "max_completion_tokens": 50,
        "temperature": 0
    }'

预期结果：

{"id":"019eab54ead036b23e53f3a709e09289","object":"chat.completion","created":1780990929,"model":"deepseek_v3.2","choices":[{"index":0,"message":{"role":"assistant","content":"The future of AI is **not a single destination, but a complex, multi-faceted trajectory** that will reshape nearly every aspect of human society, technology, and our understanding of intelligence itself. It can be understood through several interconnected lenses:\n\n### "},"finish_reason":"length"}],"usage":{"prompt_tokens":9,"completion_tokens":50,"total_tokens":59,"completion_tokens_details":{"reasoning_tokens":0},"prompt_tokens_details":{"cached_tokens":0},"prompt_cache_hit_tokens":0,"prompt_cache_miss_tokens":9},"system_fingerprint":""}

7 Accuracy Evaluation

这里提供两种精度评估方法。

使用 AISBench

1. 详细信息请参考使用 AISBench。

2. 执行后即可获取结果。

使用 Language Model Evaluation Harness

以 gsm8k 数据集作为测试数据集为例，在线模式下运行 DeepSeek-V3.2-W8A8 的精度评估。

1. 请参考使用 lm_eval 安装 lm_eval。

2. 运行 lm_eval 执行精度评估。

   lm_eval \
   --model local-completions \
   --model_args model=/root/.cache/Eco-Tech/DeepSeek-V3.2-w8a8-mtp-QuaRot,base_url=http://127.0.0.1:8000/v1/completions,tokenized_requests=False,trust_remote_code=True \
   --tasks gsm8k \
   --output_path ./
   

3. 执行后即可获取结果。

8 Performance Evaluation

使用 AISBench

详情请参考使用AISBench进行性能评估。

性能结果如下：

硬件：A3-752T，4节点

部署：1P1D，Prefill节点：DP2+TP16，Decode节点：DP8+TP4

输入/输出：64k/3k

性能：533tps，TPOT 32ms

使用vLLM基准测试

以DeepSeek-V3.2-W8A8为例进行性能评估。

更多详情请参考vllm基准测试。

vllm bench包含三个子命令：

• latency：对单批请求的延迟进行基准测试。
• serve：对在线服务吞吐量进行基准测试。
• throughput：对离线推理吞吐量进行基准测试。

以serve为例，运行代码如下。

export VLLM_USE_MODELSCOPE=True
vllm bench serve --model /root/.cache/Eco-Tech/DeepSeek-V3.2-w8a8-mtp-QuaRot  --dataset-name random --random-input 200 --num-prompts 200 --request-rate 1 --save-result --result-dir ./

9 Function call

函数调用功能从v0.13.0rc1版本开始支持，请使用最新版本。

详情请参考DeepSeek-V3.2使用指南。

10 Performance Tuning

Notice: max-model-len and max-num-seqs need to be set according to the actual usage scenario. For other settings, please refer to the deployment examples in Chapter 5.

11 FAQ

• Common Issues Tip: If you encounter issues, Refer to FAQs.

• Q: How to resolve empty output or garbled characters in PD separated deployment?

A: This is a known issue. Please ensure --no-enable-prefix-caching is set and use the latest version of vllm-ascend.

• Q: How to configure the model weight path correctly?

A: The model weight path in the deployment commands (e.g., /root/.cache/Eco-Tech/DeepSeek-V3.2-w8a8-mtp-QuaRot) should be adjusted to your actual download path. Refer to the Model Weight section for details.