概览
在 OpenCompass 中评估一个模型通常包括以下几个阶段:配置 -> 推理 -> 评估 -> 可视化。
- 配置:这是整个工作流的起点。您需要配置整个评估过程,选择要评估的模型和数据集。此外,还可以选择评估策略、计算后端等,并定义显示结果的方式。
- 推理与评估:在这个阶段,OpenCompass 将会开始对模型和数据集进行并行推理和评估。推理阶段主要是让模型从数据集产生输出,而评估阶段则是衡量这些输出与标准答案的匹配程度。这两个过程会被拆分为多个同时运行的“任务”以提高效率。
- 可视化:评估完成后,OpenCompass 将结果整理成易读的表格,并将其保存为 CSV 和 TXT 文件。
接下来,我们将展示 OpenCompass 的基础用法,分别用命令行方式和配置文件的方式评测InternLM2-Chat-1.8B,展示书生浦语在 C-Eval
基准任务上的评估。更多评测技巧请查看 快速开始 — OpenCompass 0.3.2.post1 文档 文档。
环境配置, 选择cuda12.2
安装——面向GPU的环境安装
conda create -n opencompass python=3.10
conda activate opencompass
conda install pytorch==2.1.2 torchvision==0.16.2 torchaudio==2.1.2 pytorch-cuda=12.1 -c pytorch -c nvidia -y
# 注意:一定要先 cd /root
cd /root
git clone -b 0.2.4 https://github.com/open-compass/opencompass
cd opencompass
pip install -e .
apt-get update
apt-get install cmake
pip install -r requirements.txt
pip install protobuf
安装过程1
安装过程2
安装过程3
数据准备
评测数据集
解压评测数据集到 /root/opencompass/data/
处。(注意: 上方在git clone opencompass 时一定要将 opencompass clone 到 /root 路径下)
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
InternLM和ceval 相关的配置文件
列出所有跟 InternLM 及 C-Eval 相关的配置
python tools/list_configs.py internlm ceval
运行结果:
启动评测 (10% A100 8GB 资源)
使用命令行配置参数法进行评测
打开 opencompass文件夹下configs/models/hf_internlm/的hf_internlm2_chat_1_8b.py
,贴入以下代码
from opencompass.models import HuggingFaceCausalLM
models = [
dict(
type=HuggingFaceCausalLM,
abbr='internlm2-1.8b-hf',
path="/share/new_models/Shanghai_AI_Laboratory/internlm2-chat-1_8b",
tokenizer_path='/share/new_models/Shanghai_AI_Laboratory/internlm2-chat-1_8b',
model_kwargs=dict(
trust_remote_code=True,
device_map='auto',
),
tokenizer_kwargs=dict(
padding_side='left',
truncation_side='left',
use_fast=False,
trust_remote_code=True,
),
max_out_len=100,
min_out_len=1,
max_seq_len=2048,
batch_size=8,
run_cfg=dict(num_gpus=1, num_procs=1),
)
]
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 InternLM2-Chat-1.8B 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。
#环境变量配置
export MKL_SERVICE_FORCE_INTEL=1
#或
export MKL_THREADING_LAYER=GNU
python run.py --datasets ceval_gen --models hf_internlm2_chat_1_8b --debug
运行过程中报错:
(opencompass) root@intern-studio-50008206:~/opencompass# python run.py --datasets ceval_gen --models hf_internlm2_chat_1_8b --debug
Traceback (most recent call last):
File "/root/opencompass/run.py", line 1, in <module>
from opencompass.cli.main import main
File "/root/opencompass/opencompass/cli/main.py", line 14, in <module>
from opencompass.utils.run import (exec_mm_infer_runner, fill_eval_cfg,
File "/root/opencompass/opencompass/utils/run.py", line 7, in <module>
from opencompass.datasets.custom import make_custom_dataset_config
File "/root/opencompass/opencompass/datasets/__init__.py", line 1, in <module>
from .advglue import * # noqa: F401, F403
File "/root/opencompass/opencompass/datasets/advglue.py", line 4, in <module>
from datasets import Dataset, concatenate_datasets
File "/root/.conda/envs/opencompass/lib/python3.10/site-packages/datasets/__init__.py", line 18, in <module>
from .arrow_dataset import Dataset
File "/root/.conda/envs/opencompass/lib/python3.10/site-packages/datasets/arrow_dataset.py", line 59, in <module>
import pandas as pd
File "/root/.conda/envs/opencompass/lib/python3.10/site-packages/pandas/__init__.py", line 22, in <module>
from pandas.compat import is_numpy_dev as _is_numpy_dev # pyright: ignore # noqa:F401
File "/root/.conda/envs/opencompass/lib/python3.10/site-packages/pandas/compat/__init__.py", line 18, in <module>
from pandas.compat.numpy import (
File "/root/.conda/envs/opencompass/lib/python3.10/site-packages/pandas/compat/numpy/__init__.py", line 4, in <module>
from pandas.util.version import Version
File "/root/.conda/envs/opencompass/lib/python3.10/site-packages/pandas/util/__init__.py", line 2, in <module>
from pandas.util._decorators import ( # noqa:F401
File "/root/.conda/envs/opencompass/lib/python3.10/site-packages/pandas/util/_decorators.py", line 14, in <module>
from pandas._libs.properties import cache_readonly
File "/root/.conda/envs/opencompass/lib/python3.10/site-packages/pandas/_libs/__init__.py", line 13, in <module>
from pandas._libs.interval import Interval
File "pandas/_libs/interval.pyx", line 1, in init pandas._libs.interval
ValueError: numpy.dtype size changed, may indicate binary incompatibility. Expected 96 from C header, got 88 from PyObject
重新安装:
pip uninstall numpy
pip install numpy==1.23.5
运行结果:
经过4个小时测评,评测完成后,将会看到:图1
图2
图3
使用配置文件修改参数法进行评测
除了通过命令行配置实验外,OpenCompass 还允许用户在配置文件中编写实验的完整配置,并通过 run.py 直接运行它。配置文件是以 Python 格式组织的,并且必须包括 datasets 和 models 字段。本次测试配置在 configs
文件夹 中。此配置通过 继承机制 引入所需的数据集和模型配置,并以所需格式组合 datasets 和 models 字段。 运行以下代码,在configs文件夹下创建eval_tutorial_demo.py
cd /root/opencompass/configs
touch eval_tutorial_demo.py
打开eval_tutorial_demo.py
贴入以下代码
from mmengine.config import read_base
with read_base():
from .datasets.ceval.ceval_gen import ceval_datasets
from .models.hf_internlm.hf_internlm2_chat_1_8b import models as hf_internlm2_chat_1_8b_models
datasets = ceval_datasets
models = hf_internlm2_chat_1_8b_models
因此,运行任务时,我们只需将配置文件的路径传递给 run.py:
cd /root/opencompass
python run.py configs/eval_tutorial_demo.py --debug
如果一切正常,您应该看到屏幕上显示:
经过4个小时测评,评测完成后,将会看到:
图1
图2
图3