Gaokao and some datasets appear many zero when I evaluate them [Bug]

Prerequisite

• I have searched Issues and Discussions but cannot get the expected help.
• The bug has not been fixed in the latest version.

Type

I'm evaluating with the officially supported tasks/models/datasets.

Environment

latest environment

Reproduces the problem - code/configuration sample

1

Reproduces the problem - command or script

CUDA_VISIBLE_DEVICES='0,1,2,3,4,5,6,7' python run.py --hf-path /Llama-2-7b-hf --datasets gsm8k_gen_1d7fe4 bbh_gen math_gen_265cce GaokaoBench_gen_5cfe9e agieval_gen_a0c741 --model-kwargs device_map='auto' --tokenizer-kwargs padding_side='left' truncation='left' use_fast=False --max-out-len 100 --max-seq-len 4096 --batch-size 8 --no-batch-padding --num-gpus 1 --max-partition-size 15000

Reproduces the problem - error message

appear too much zero ;
and in my evaluation, result of Gaokao for llmam2-7b-hf is 7.06 which is not consistent with 18.9 in https://opencompass.org.cn/leaderboard-llm
thank u and much appreciate

Other information

i want to reproduce 18.9 in your website !!