第7节 OpenCompass 大模型评测实战 作业

基础作业

• 使用 OpenCompass 评测 internlm2-chat-1_8b 模型在 C-Eval 数据集上的性能

安装

studio-conda -o internlm-base -t opencompass
source activate opencompass
git clone -b 0.2.4 https://github.com/open-compass/opencompass
cd opencompass
pip install -e .

pip install -r requirements.txt

数据准备

cd /root
mkdir opencompass
cp /share/temp/datasets/OpenCompassData-core-20231110.zip /opencompass
cd /opencompass
unzip OpenCompassData-core-20231110.zip

python tools/list_configs.py internlm ceval

启动评测 (10% A100 8GB 资源)

先安装依赖，配置环境变量，避免报错

pip install -r requirements.txt

pip install protobuf

export MKL_SERVICE_FORCE_INTEL=1
#或
export MKL_THREADING_LAYER=GNU

python run.py --datasets ceval_gen --hf-path /share/new_models/Shanghai_AI_Laboratory/internlm2-chat-1_8b --tokenizer-path /share/new_models/Shanghai_AI_Laboratory/internlm2-chat-1_8b --tokenizer-kwargs padding_side='left' truncation='left' trust_remote_code=True --model-kwargs trust_remote_code=True device_map='auto' --max-seq-len 1024 --max-out-len 16 --batch-size 2 --num-gpus 1 --debug