Files
enginex-ascend-910-vllm/tests/e2e/nightly/single_node/models/configs/Qwen3.5-27B-w8a8-A2.yaml
Sun Ruoxi 7f8a1b1f7a init v0.23.0
Signed-off-by: Sun Ruoxi <sunruoxi@4paradigm.com>
2026-08-27 15:11:51 +08:00

58 lines
1.8 KiB
YAML

# ==========================================
# ACTUAL TEST CASES
# ==========================================
test_cases:
- name: "Qwen3.5-27B-w8a8"
model: "Eco-Tech/Qwen3.5-27B-w8a8-mtp"
envs:
VLLM_USE_MODELSCOPE: "true"
PYTORCH_NPU_ALLOC_CONF: "expandable_segments:True"
HCCL_BUFFSIZE: "1024"
HCCL_OP_EXPANSION_MODE: "AIV"
OMP_NUM_THREADS: "1"
TASK_QUEUE_ENABLE: "1"
VLLM_ASCEND_ENABLE_PREFETCH_MLP: "1"
VLLM_ASCEND_ENABLE_DENSE_OPTIMIZE: "1"
VLLM_ASCEND_ENABLE_NZ: "1"
VLLM_ASCEND_ENABLE_FUSED_MC2: "1"
SERVER_PORT: "DEFAULT_PORT"
server_cmd:
- "--tensor-parallel-size"
- "2"
- "--port"
- "$SERVER_PORT"
- "--max-num-seqs"
- "128"
- "--quantization"
- "ascend"
- "--max-model-len"
- "262144"
- "--max-num-batched-tokens"
- "8192"
- "--trust-remote-code"
- "--gpu-memory-utilization"
- "0.95"
- "--additional-config"
- '{"enable_cpu_binding":true, "enable_weight_nz_layout":true}'
- "--speculative_config"
- '{"method": "qwen3_5_mtp", "num_speculative_tokens": 3,"enforce_eager": true}'
- "--compilation-config"
- '{"cudagraph_mode":"FULL_DECODE_ONLY", "cudagraph_capture_sizes":[4,8,12,16,20,24,28,32,36,40,44,48,52,56,60,64,68,72,76,80,84,88,92,96,100,104,108,112,116,120,124,128,132,136,140,144]}'
- "--mm-processor-cache-gb"
- "0"
- "--mm_processor_cache_type"
- "shm"
benchmarks:
perf:
case_type: performance
dataset_path: vllm-ascend/GSM8K-in3500-bs8000-qwen3
request_conf: vllm_api_stream_chat
dataset_conf: gsm8k/gsm8k_gen_0_shot_cot_str_perf
num_prompts: 128
max_out_len: 1500
batch_size: 32
request_rate: 0
baseline: 604
threshold: 0.97