MLP Tabular Regression

This example trains an MLP on the UCI Auto MPG regression task. The prepared CSV has seven normalized numeric car features and one normalized target column for miles per gallon, so the model is just ordinary supervised tabular regression:

x1..x7 -> Linear -> ReLU -> Linear -> y.

Prepare the CSV once:

python3 scripts/datasets/download_example_data.py --auto-mpg
lake exe torchlean mlp --cpu --steps 1

The downloader writes normalized columns x1..x7,y. If you want to try your own tabular regression CSV, pass --csv PATH with the same columns.

def NN.Examples.Models.Supervised.Mlp.exeName : String

CLI subcommand name used in terminal banners and error messages.

def NN.Examples.Models.Supervised.Mlp.defaultLogJson : System.FilePath

Default JSON loss-curve path for this command.

def NN.Examples.Models.Supervised.Mlp.batch : ℕ

Static minibatch size for the Auto MPG tabular loader.

def NN.Examples.Models.Supervised.Mlp.inDim : ℕ

Auto MPG has seven numeric predictors after dropping car_name.

def NN.Examples.Models.Supervised.Mlp.hidDim : ℕ

Hidden width of the one-hidden-layer MLP.

def NN.Examples.Models.Supervised.Mlp.outDim : ℕ

Regression target width: normalized miles-per-gallon.

def NN.Examples.Models.Supervised.Mlp.cfg : TorchLean.nn.models.Mlp1Config

Shared MLP configuration used by shapes and the constructor.

@[reducible, inline]

abbrev NN.Examples.Models.Supervised.Mlp.σ : Spec.Shape

Input shape: a minibatch of Auto MPG feature vectors.

@[reducible, inline]

abbrev NN.Examples.Models.Supervised.Mlp.τ : Spec.Shape

Output shape: one scalar regression prediction per row.

def NN.Examples.Models.Supervised.Mlp.model : TorchLean.nn.M (TorchLean.nn.Sequential σ τ)

One-hidden-layer ReLU MLP from the public model API.

def NN.Examples.Models.Supervised.Mlp.data (path : System.FilePath) (seed : ℕ) : TorchLean.Trainer.Dataset σ τ

Auto MPG as a public TorchLean dataset.

The only dataset-specific details here are the CSV path, header convention, batch size, and feature count. Runtime scalar selection stays inside Trainer, so the same dataset works for CPU, CUDA, compiled, eager, and checked scalar modes.

def NN.Examples.Models.Supervised.Mlp.train (opts : TorchLean.Options) (flags : TorchLean.ModelZoo.CsvTrainFlags) : IO (TorchLean.Trainer.TrainResult σ τ)

Train the Auto MPG MLP with the public Trainer surface.

def NN.Examples.Models.Supervised.Mlp.main (args : List String) : IO UInt32

CLI entrypoint for Auto MPG regression on CPU or CUDA.