OpenCompass/configs/models/hf_llama/lmdeploy_llama2_13b_chat.py

16 lines
473 B
Python

from opencompass.models import TurboMindModelwithChatTemplate
models = [
dict(
type=TurboMindModelwithChatTemplate,
abbr='llama-2-13b-chat-turbomind',
path='meta-llama/Llama-2-13b-chat-hf',
engine_config=dict(max_batch_size=16, tp=1),
gen_config=dict(top_k=1, temperature=1e-6, top_p=0.9, max_new_tokens=1024),
max_seq_len=4096,
max_out_len=1024,
batch_size=16,
run_cfg=dict(num_gpus=1),
)
]