OpenCompass/opencompass/configs/models/qwen3/lmdeploy_qwen3_0_6b.py
2025-05-27 19:41:13 +08:00

19 lines
670 B
Python

from opencompass.models import TurboMindModelwithChatTemplate
from opencompass.utils.text_postprocessors import extract_non_reasoning_content
model = [
dict(
type=TurboMindModelwithChatTemplate,
abbr='qwen_3_0.6b_thinking-turbomind',
path='Qwen/Qwen3-0.6B',
engine_config=dict(session_len=32768, max_batch_size=16, tp=1),
gen_config=dict(
top_k=20, temperature=0.6, top_p=0.95, do_sample=True, enable_thinking=True
),
max_seq_len=32768,
max_out_len=32000,
batch_size=16,
run_cfg=dict(num_gpus=1),
pred_postprocessor=dict(type=extract_non_reasoning_content)
),
]