Add a conda environment

This page describes how to add a conda environment to your Vertex AI Workbench instance.

Overview

When you add a conda environment to your Vertex AI Workbench instance, it appears as a kernel in your instance's JupyterLab interface.

You might add a conda environment to your Vertex AI Workbench instance to use kernels that aren't available in Vertex AI Workbench instances. For example, you can add conda environments for R and Apache Beam. Or you can add conda environments for specific older versions of the available frameworks, such as TensorFlow, PyTorch, or Python.

Before you begin

If you haven't already, create a Vertex AI Workbench instance.

Open JupyterLab

  1. In the Google Cloud console, go to the Instances page.

    Go to Instances

  2. Next to your Vertex AI Workbench instance's name, click Open JupyterLab.

    Your Vertex AI Workbench instance opens JupyterLab.

Add a conda environment

You can add a conda environment by entering commands in your instance's JupyterLab terminal.

  1. In JupyterLab, select File > New > Terminal.

  2. In the Terminal window, enter the following commands:

    
# Creates a conda environment.
conda create -n CONDA_ENVIRONMENT_NAME -y
conda activate CONDA_ENVIRONMENT_NAME

# Install packages using a pip local to the conda environment.
conda install pip
pip install PACKAGE

# Adds the conda kernel.
DL_ANACONDA_ENV_HOME="${DL_ANACONDA_HOME}/envs/CONDA_ENVIRONMENT_NAME"
python -m ipykernel install --prefix "${DL_ANACONDA_ENV_HOME}" --name CONDA_ENVIRONMENT_NAME --display-name KERNEL_DISPLAY_NAME

    Replace the following:

    • CONDA_ENVIRONMENT_NAME: your choice of name for the environment
    • PACKAGE: the conda package that you want to install
    • KERNEL_DISPLAY_NAME: the display name for the tile of the kernel in the JupyterLab interface

  3. To see your new kernel, do the following:

    1. Refresh the page.

    2. Select File > New Launcher.

    The kernel is listed among the others in the Launcher window.

By default, conda might use pip packages in the system pip folder (for example, /usr/bin/pip). Running conda install pip ensures that the setup uses a pip local to the environment.

Example installation: R Essentials

The following example installs R Essentials in a conda environment named r.

conda create -n r
conda activate r
conda install -c r r-essentials

DL_ANACONDA_ENV_HOME="${DL_ANACONDA_HOME}/envs/r"
python -m ipykernel install --prefix "${DL_ANACONDA_ENV_HOME}" --name r --display-name r
rm -rf /opt/conda/envs/r/share/jupyter/kernels/python3
rm -rf /opt/conda/envs/r/share/jupyter/kernels/ir

Example installation: pip package

The following example installs pip packages from a requirements.txt file.

conda create -n myenv
conda activate myenv
conda install pip
pip install -r requirements.txt
DL_ANACONDA_ENV_HOME="${DL_ANACONDA_HOME}/envs/myenv"
python -m ipykernel install --prefix "${DL_ANACONDA_ENV_HOME}" --name myenv --display-name myenv

Troubleshoot

To diagnose and resolve issues related to adding a conda environment, see Troubleshooting Vertex AI Workbench.

What's next