Quick start¶

This page gives the shortest path from cloning the repository to seeing a tracked experiment in Aim.

The commands below run natively on Linux and macOS. On Windows, run them inside WSL2 because Aim does not support native Windows.

Clone the repository¶

git clone https://github.com/CosmosRedshift7/ml-template.git
cd ml-template

pixi install

This creates a local Pixi environment from pyproject.toml and pixi.lock.

pixi run train

This trains the example model and writes local outputs under:

local/

Typical outputs include:

local/aim/
local/checkpoints/
local/figures/

pixi run aim-ui

Then open:

http://127.0.0.1:43800

In Aim, open the ml-template experiment to view metrics, parameters, and tracked figures.

After training, evaluate the best checkpoint:

pixi run evaluate

By default, this evaluates:

local/checkpoints/best.ckpt

Before editing the template, run:

pixi run pytest

This confirms that the project imports correctly and the smoke tests pass.