5 minute setup

Quickstart

Get started with mlcli in under 5 minutes. This guide will walk you through installation and your first training run.

Installation

Install mlcli using pip:

Terminal
pip install mlcli-toolkit

Or using pipx for an isolated installation:

Terminal
pipx install mlcli-toolkit

Verify the installation:

Terminal
mlcli --version
# mlcli v0.3.1

Your First Training Run

Train a Random Forest model on your dataset with a single command:

Terminal
mlcli train \
  --data data/train.csv \
  --model random_forest \
  --target label \
  --output models/

This command will:

Load and preprocess your data
Train a Random Forest classifier
Evaluate on a validation split
Save the trained model
Log the experiment

Using Config Files

For more control over your training, use a JSON configuration file:

config.json
{
  "model_type": "random_forest",
  "dataset_path": "data/train.csv",
  "target_column": "label",
  "test_size": 0.2,
  "hyperparameters": {
    "n_estimators": 100,
    "max_depth": 10,
    "min_samples_split": 2
  },
  "output_dir": "models/"
}

Then run training with the config:

Terminal
mlcli train --config config.json

Evaluate Your Model

Evaluate your trained model on a test set:

Terminal
mlcli eval \
  --model models/random_forest_model.pkl \
  --data data/test.csv

View Experiments

List all your experiment runs:

Terminal
mlcli list-runs

View details of a specific run:

Terminal
mlcli show-run --run-id run_abc123

Next Steps

Explore TrainersConfiguration Guide