Files
enginex-ascend-910-vllm/tests/e2e/nightly/multi_node/external_dp/config/DeepSeek-V4-flash-w8a8-PD-prefix.yaml
Sun Ruoxi 7f8a1b1f7a init v0.23.0
Signed-off-by: Sun Ruoxi <sunruoxi@4paradigm.com>
2026-08-27 15:11:51 +08:00

177 lines
4.7 KiB
YAML

test_name: "DeepSeek-V4-Flash-w8a8-PD-prefix"
model: "Eco-Tech/DeepSeek-V4-Flash-w8a8-mtp"
num_nodes: 2
npu_per_node: 16
routing:
type: "disaggregated_prefill"
groups:
prefiller: [ 0 ]
decoder: [ 1 ]
config:
- node_index: 0
port_start: 7100
dp_rpc_port: 12321
dp_size: 4
dp_size_local: 4
dp_rank_start: 0
tp_size: 4
dp_address: "${NODE_0_IP}"
- node_index: 1
port_start: 7100
dp_rpc_port: 12321
dp_size: 16
dp_size_local: 16
dp_rank_start: 0
tp_size: 1
dp_address: "${NODE_1_IP}"
env_common: &env_common
VLLM_USE_MODELSCOPE: "true"
SERVER_PORT: "${PORT}"
ASCEND_RT_VISIBLE_DEVICES: "${VISIBLE_DEVICES}"
VLLM_RPC_TIMEOUT: "3600"
VLLM_EXECUTE_MODEL_TIMEOUT_SECONDS: "30000"
HCCL_EXEC_TIMEOUT: "204"
HCCL_CONNECT_TIMEOUT: "120"
OMP_PROC_BIND: "false"
OMP_NUM_THREADS: "10"
PYTORCH_NPU_ALLOC_CONF: "expandable_segments:True"
env_prefill: &env_prefill
<<: *env_common
HCCL_BUFFSIZE: "1024"
TASK_QUEUE_ENABLE: "1"
HCCL_OP_EXPANSION_MODE: "AIV"
VLLM_ASCEND_ENABLE_FUSED_MC2: "1"
VLLM_ASCEND_ENABLE_FLASHCOMM1: "1"
VLLM_PREFIX_CACHE_RETENTION_INTERVAL: "4096"
env_decode: &env_decode
<<: *env_common
HCCL_BUFFSIZE: "1500"
templates:
- node_index: 0
envs:
<<: *env_prefill
server_cmd_template:
- --host
- "0.0.0.0"
- --port
- $SERVER_PORT
- --data-parallel-size
- ${DP_SIZE}
- --data-parallel-rank
- ${DP_RANK}
- --data-parallel-address
- ${DP_ADDRESS}
- --data-parallel-rpc-port
- ${DP_RPC_PORT}
- --tensor-parallel-size
- ${TP_SIZE}
- --enable-expert-parallel
- --seed
- "1024"
- --max-model-len
- "1048576"
- --max-num-batched-tokens
- "8192"
- --max-num-seqs
- "16"
- --no-disable-hybrid-kv-cache-manager
- --model-loader-extra-config
- '{"enable_multithread_load": true, "num_threads": 128}'
- --speculative-config
- '{"num_speculative_tokens": 1,"method": "mtp"}'
- --trust-remote-code
- --block-size
- "32"
- --tokenizer-mode
- "deepseek_v4"
- --tool-call-parser
- "deepseek_v4"
- --enable-auto-tool-choice
- --reasoning-parser
- "deepseek_v4"
- --gpu-memory-utilization
- "0.9"
- --quantization
- "ascend"
- --enforce-eager
- --additional-config
- '{"enable_cpu_binding": true, "enable_dsa_cp": true}'
- --kv-transfer-config
- '{"kv_connector": "MooncakeHybridConnector", "kv_role": "kv_producer", "kv_port": "30000", "engine_id": "0", "kv_connector_extra_config": {"prefill": {"dp_size": 4, "tp_size": 4}, "decode": {"dp_size": 16, "tp_size": 1}}}'
- node_index: 1
envs:
<<: *env_decode
server_cmd_template:
- --host
- "0.0.0.0"
- --port
- $SERVER_PORT
- --data-parallel-size
- ${DP_SIZE}
- --data-parallel-rank
- ${DP_RANK}
- --data-parallel-address
- ${DP_ADDRESS}
- --data-parallel-rpc-port
- ${DP_RPC_PORT}
- --tensor-parallel-size
- ${TP_SIZE}
- --enable-expert-parallel
- --seed
- "1024"
- --max-model-len
- "1048576"
- --max-num-batched-tokens
- "240"
- --max-num-seqs
- "60"
- --async-scheduling
- --block-size
- "32"
- --no-enable-prefix-caching
- --no-disable-hybrid-kv-cache-manager
- --model-loader-extra-config
- '{"enable_multithread_load": true, "num_threads": 128}'
- --trust-remote-code
- --tokenizer-mode
- "deepseek_v4"
- --tool-call-parser
- "deepseek_v4"
- --enable-auto-tool-choice
- --reasoning-parser
- "deepseek_v4"
- --gpu-memory-utilization
- "0.95"
- --quantization
- "ascend"
- --speculative-config
- '{"num_speculative_tokens": 3,"method": "mtp"}'
- --compilation-config
- '{"cudagraph_mode": "FULL_DECODE_ONLY"}'
- --kv-transfer-config
- '{"kv_connector": "MooncakeHybridConnector", "kv_role": "kv_consumer", "kv_port": "30100", "engine_id": "1", "kv_connector_extra_config": {"prefill": {"dp_size": 4, "tp_size": 4}, "decode": {"dp_size": 16, "tp_size": 1}}}'
- --additional-config
- '{"ascend_compilation_config":{"enable_npugraph_ex":true,"enable_static_kernel":false},"enable_cpu_binding":true,"multistream_overlap_shared_expert":true,"recompute_scheduler_enable":true}'
benchmarks:
acc_aime2025:
case_type: accuracy
dataset_path: vllm-ascend/aime2025
request_conf: vllm_api_general_chat
dataset_conf: aime2025/aime2025_gen_0_shot_chat_prompt
temperature: 1.0
top_p: 1.0
thinking: "true"
max_out_len: 65536
batch_size: 32
baseline: 90
threshold: 10