18 lines
397 B
YAML
18 lines
397 B
YAML
model_name: "Qwen/Qwen3-30B-A3B"
|
|
tasks:
|
|
- name: "gsm8k"
|
|
metrics:
|
|
- name: "exact_match,strict-match"
|
|
value: 0.89
|
|
- name: "exact_match,flexible-extract"
|
|
value: 0.85
|
|
- name: "ceval-valid"
|
|
metrics:
|
|
- name: "acc,none"
|
|
value: 0.84
|
|
num_fewshot: 5
|
|
gpu_memory_utilization: 0.6
|
|
enable_expert_parallel: True
|
|
tensor_parallel_size: 2
|
|
apply_chat_template: False
|
|
fewshot_as_multiturn: False |