valohai-cli

Valohai Command Line Client

This is the command-line client for the Valohai machine learning IaaS platform.

Installation

valohai-cli supports Python 3.8 and higher.

If you still need to run on Python 3.5, version 0.13.0 was the last one to support it. If you still need to run on Python 3.6, version 0.23.0 was the last one to support it. If you still need to run on Python 3.7, version 0.25.0 was the last one to support it.

System-wide or user-wide installation with pipx

The recommended way to install valohai-cli system-wide is to use pipx, an utility to install and run Python applications in isolated environments. (If you're familiar with Node.js's npx tool, it's the same idea.)

This ensures that valohai-cli's dependencies don't conflict with other Python packages.

Once you have installed and configured pipx (see the link above), you can

$ pipx install valohai-cli

and to upgrade it later on,

$ pipx upgrade valohai-cli

System-wide or user-wide installation with pip

You can also install valohai-cli system-wide with pip, but this may cause conflicts with other Python packages installed system-wide or user-wide.

$ pip3 install -U valohai-cli

The -U flag ensures that any present version is upgraded, too.

Installation in a virtual environment

If you prefer to install valohai-cli in a virtual environment, you can do so with pip as well.

$ python3 -m venv venv
$ source venv/bin/activate
$ pip install -U valohai-cli

Getting Started

After you've installed the client, vh should work and you should see a description of commands.

See the tutorial document!

Developing

Development requires Python 3.10+; otherwise you'll get false positive type failures. CI will run tests on older Python versions.

To work on the valohai-cli code: pull the repository, create and activate a virtualenv, then run:

make dev

This installs valohai-cli as an "editable" vh command available in the virtualenv, but linked to the working copy's source. That is, you can now edit the source under valohai_cli in your working directory, and try it out with vh.

vh --help
# Usage: vh [OPTIONS] COMMAND [ARGS]...

To run lints, type checks and tests:

# run linting and type checks
make lint

# run tests
make test