Generated Tutorial Dataset Paths #
TorchLean tutorials can generate deterministic sample datasets under NN/Examples/Data/.
This module centralizes:
- the default data directory,
- the file paths of the generated sample datasets, and
- a CLI helper for overriding the directory via
--data-dir.
Keeping these paths in one module keeps tutorial code from hardcoding filenames or reimplementing the same flag parsing in each example.
Default relative directory containing generated tutorial datasets.
Instances For
Parse an optional --data-dir PATH flag (defaults to defaultDataDir).
Tutorials use this when they load generated CSV/NPY files from disk.
Instances For
Resolved --x / --y dataset paths for sample-data tutorials.
- xPath : System.FilePath
Feature or image tensor path.
- yPath : System.FilePath
Label or target tensor path.
Instances For
Parse optional --x / --y overrides and fall back to the supplied sample-data defaults.
Tutorials that load paired .npy files use this instead of re-implementing the same
takePathFlagOnce / getD boilerplate in every example.
Instances For
small_regression.csv (2D regression).
Instances For
small_regression_X.npy (shape 25x2).
Instances For
small_regression_y.npy (shape 25x1).
Instances For
small_cifar10like_X.npy (shape 200x3x32x32).
Instances For
small_cifar10like_y.npy (shape 200).