OpenCompass 评测 InternLM-1.8B 实践-材料提交

安装

面向GPU的环境安装

conda create --name opencompass --clone=/root/share/conda_envs/internlm-base
source activate opencompass
git clone https://github.com/open-compass/opencompass
cd opencompass
pip install -e .

数据准备

# 解压评测数据集到 data/ 处
cp /share/temp/datasets/OpenCompassData-core-20231110.zip /root/opencompass/
unzip OpenCompassData-core-20231110.zip

# 将会在opencompass下看到data文件夹

查看支持的数据集和模型

# 列出所有跟 internlm 及 ceval 相关的配置
python tools/list_configs.py internlm ceval

+--------------------------+--------------------------------------------------------+
| Model                    | Config Path                                            |
|--------------------------+--------------------------------------------------------|
| hf_internlm_20b          | configs/models/hf_internlm/hf_internlm_20b.py          |
| hf_internlm_7b           | configs/models/hf_internlm/hf_internlm_7b.py           |
| hf_internlm_chat_20b     | configs/models/hf_internlm/hf_internlm_chat_20b.py     |
| hf_internlm_chat_7b      | configs/models/hf_internlm/hf_internlm_chat_7b.py      |
| hf_internlm_chat_7b_8k   | configs/models/hf_internlm/hf_internlm_chat_7b_8k.py   |
| hf_internlm_chat_7b_v1_1 | configs/models/hf_internlm/hf_internlm_chat_7b_v1_1.py |
| internlm_7b              | configs/models/internlm/internlm_7b.py                 |
| ms_internlm_chat_7b_8k   | configs/models/ms_internlm/ms_internlm_chat_7b_8k.py   |
+--------------------------+--------------------------------------------------------+
+----------------------------+------------------------------------------------------+
| Dataset                    | Config Path                                          |
|----------------------------+------------------------------------------------------|
| ceval_clean_ppl            | configs/datasets/ceval/ceval_clean_ppl.py            |
| ceval_gen                  | configs/datasets/ceval/ceval_gen.py                  |
| ceval_gen_2daf24           | configs/datasets/ceval/ceval_gen_2daf24.py           |
| ceval_gen_5f30c7           | configs/datasets/ceval/ceval_gen_5f30c7.py           |
| ceval_ppl                  | configs/datasets/ceval/ceval_ppl.py                  |
| ceval_ppl_578f8d           | configs/datasets/ceval/ceval_ppl_578f8d.py           |
| ceval_ppl_93e5ce           | configs/datasets/ceval/ceval_ppl_93e5ce.py           |
| ceval_zero_shot_gen_bd40ef | configs/datasets/ceval/ceval_zero_shot_gen_bd40ef.py |
+----------------------------+------------------------------------------------------+

启动评测

确保按照上述步骤正确安装 OpenCompass 并准备好数据集后，可以通过以下命令评测 InternLM-Chat-7B 模型在 C-Eval 数据集上的性能。由于 OpenCompass 默认并行启动评估过程，我们可以在第一次运行时以 --debug 模式启动评估，并检查是否存在问题。在 --debug 模式下，任务将按顺序执行，并实时打印输出。

python run.py --datasets ceval_gen --hf-path /share/temp/model_repos/internlm-chat-7b/ --tokenizer-path /share/temp/model_repos/internlm-chat-7b/ --tokenizer-kwargs padding_side='left' truncation='left' trust_remote_code=True --model-kwargs trust_remote_code=True device_map='auto' --max-seq-len 2048 --max-out-len 16 --batch-size 4 --num-gpus 1 --debug

 评测完成后，将会看到：