Malet

Machine Learning Experiment Tool

Malet is a tool for hyperparameter grid searches, metric logging, advanced analyses and visualizations.

🔎 Easy & powerful grid search

Define hyperparameter grids in a single YAML file. Malet handles the combinatorial expansion, parallel splitting, resuming, and logging automatically.

📝 Experiment logging & resuming

Metric logging with automatic TSV persistence and experiment resuming — pick up where you left off when jobs are killed.

📊 Flexible visualization

Generate curves, bar charts, heatmaps, scatter plots, and animated GIFs directly from experiment logs with malet-plot. No boilerplate scripts needed.

🚀 Multi-GPU parallelization

Split experiment grids across GPUs with Slurm, checkpoint long-running jobs, and merge results — all with built-in file locking.

---

Getting started

Install Malet and run your first experiment in minutes:

pip install malet

# Run experiments over a grid of hyperparameters
python train.py

# Plot results with a single command
malet-plot -exp_folder ./experiments/my_exp -mode curve-epoch-val_accuracy

New to Malet? Start with the Installation guide, then work through the Quick Start.