Files
enginex-ascend-910-vllm/tests/e2e/models/configs/Qwen3-Next-80B-A3B-Instruct.yaml
Sun Ruoxi 7f8a1b1f7a init v0.23.0
Signed-off-by: Sun Ruoxi <sunruoxi@4paradigm.com>
2026-08-27 15:11:51 +08:00

24 lines
446 B
YAML

model_name: "Qwen/Qwen3-Next-80B-A3B-Instruct"
model_type: "vllm"
hardware: "Atlas A2 Series"
serve:
tensor_parallel_size: 4
dtype: auto
max_model_len: 4096
gpu_memory_utilization: 0.7
trust_remote_code: false
enable_expert_parallel: true
enforce_eager: true
tasks:
- name: "ceval-valid_accountant"
metrics:
- name: "acc,none"
value: 0.98
num_fewshot: 5
apply_chat_template: true
fewshot_as_multiturn: true
batch_size: 1