evalstone/evalscope/docs/en/get_started/installation.md

Installation

Method 1. Installing via pip

We recommend using conda to manage your environment and pip to install dependencies. You can install the latest evalscope package from PyPI.

1. Create a conda environment (optional)

# Python 3.10 is recommended
conda create -n evalscope python=3.10

# Activate the conda environment
conda activate evalscope

2. Install dependencies via pip

pip install evalscope

3. Install additional dependencies (optional)

• To use the model service inference performance testing feature, install the perf dependency:

  pip install 'evalscope[perf]'
  

• To use visualization features, install the app dependency:

  pip install 'evalscope[service]'
  

• If you want to use other evaluation backends, you can optionally install OpenCompass, VLMEvalKit, and RAGEval:

  pip install 'evalscope[opencompass]'
  pip install 'evalscope[vlmeval]'
  pip install 'evalscope[rag]'
  

• To install all dependencies:

  pip install 'evalscope[all]'
  

Since the project has been renamed to `evalscope`, for version `v0.4.3` or earlier, you can install using:
```shell
 pip install llmuses<=0.4.3
```
To import dependencies from `llmuses`:
``` python
from llmuses import ...
```

Method 2. Installing from source

By installing from source, you can use the latest code and conveniently perform secondary development and debugging.

1. Download the source code

git clone https://github.com/modelscope/evalscope.git

2. Install dependencies

cd evalscope/

pip install -e .

3. Install additional dependencies

• To use the model service inference performance testing feature, install the perf dependency:

  pip install '.[perf]'
  

• To use visualization features, install the app dependency:

  pip install '.[app]'
  

• If you want to use other evaluation backends, you can optionally install OpenCompass, VLMEvalKit, and RAGEval:

  pip install '.[opencompass]'
  pip install '.[vlmeval]'
  pip install '.[rag]'
  

• To install all dependencies:

  pip install '.[all]'
  

Docker Image

You can use the official ModelScope Docker images, which include the EvalScope library. Refer to here for more information.