Experiment DSL

Experiment DSL#

Use MLX::DSL.experiment for declarative run wiring across model, optimizer, trainer, data, and artifacts.

exp = MLX::DSL.experiment("demo") do
  model { model }
  optimizer { optimizer }
end

Entry point#

  • MLX::DSL.experiment(name = nil) { ... }

exp = MLX::DSL.experiment("classifier") do
end

Declaration sections#

  • model { ... }

  • optimizer { ... }

  • trainer(**kwargs) { loss } or trainer(existing_trainer)

  • data(train:, validation:, **fit_kwargs)

  • artifacts(**fit_kwargs)

exp = MLX::DSL.experiment("classifier") do
  model { Classifier.new(input_dim: 128, num_classes: 10) }
  optimizer { MLX::Optimizers::Adam.new(learning_rate: 1e-3) }
  trainer { |x:, y:| MLX::NN.cross_entropy(model.call(x), y, reduction: "mean") }
  data train: train_data, validation: validation_data
  artifacts checkpoint_path: "checkpoints/exp-%{epoch}.bin"
end

Execution helpers#

  • run(report: false, **overrides)

  • report(**overrides)

  • save_run_bundle(path, report:, config:, **overrides)

exp = MLX::DSL.experiment("mnist") do
  model { model }
  optimizer { optimizer }
  trainer { |x:, y:| MLX::NN.cross_entropy(model.call(x), y, reduction: "mean") }
  data train: train_data, validation: validation_data
  artifacts checkpoint_path: "checkpoints/ep-%{epoch}.bin"
end

report = exp.report(epochs: 5)

See implementation:

  • lib/mlx/dsl/experiment.rb

Contents