How do I Use Anacoda/Cona with Inference.ai?

Written By chloe kwok

Last updated 8 months ago

Yes! Inference.ai fully supports Anaconda and conda environments.

You can use conda to manage packages, create virtual environments, and install dependencies just like you would on a local machine.

Getting Started

Once your server is running:

  1. Open a terminal (via SSH or JupyterLab)

  2. Create a new environment

Example
conda create -n myenv python=3.10
conda activate myenv

  1. Install packages as needed:

Example
conda install numpy pandas scikit-learn

Pro tip: Environments and packages are reset when the server stops, unless you are using persistent storage

Keeping Your Python Environment in Permanent Storage

Note that:

  • Conda must be installed and be on your path

  • The OS which the environment was built must match the OS of the target

  • Once an environment is unpacked and conda-unpack has been executed, it cannot be relocated

  1. Create a Server

  1. Select Show Details

  1. Select the Jupyter-labs link from the HTTP Ports segment

  1. Select Terminal

  1. Run conda create -p /tmp/env python=3.12 numpy pandas -y

  2. Run conda install -n base -c conda-forge conda-pack -y

  1. Run conda-pack -p /tmp/env -o myenv.tar.gz

  1. Run mv myenv.tar.gz /home/jovyan/work

Do this portion on the target machine

  1. Unpack the environment into your directory with $ mkdir -p my_env $ tar -xzf out_name.tar.gz -C my_env

  2. Use Python without activating or fixing the prefixes ./my_env/bin/python

  3. Activate the environment $ source my_env/bin/activate

  4. Run Python from in the environment (my_env) $ python

  5. Cleanup prefixes from in the active environment (my_env) $ conda-unpack

  6. Run (my_env) $ ipython --version

  7. Deactivate the environment to remove it from your path (my_env) $ source my_env/bin/deactivate