Splits your dataset folder into train/validation/test subsets. Very useful for non-csv data, like images.
(sudo) pip install dataset-split
dataset-split <path> (optional arguments)
| Options | Usage | Description |
|---|---|---|
| ratio | (-r or ratio) <train> <test> <valid> |
Sets the ratios that will split the dataset! Floats are interpreted as percentages. The order must be ratio-valid-test and must sum up to 1.0 (won't run otherwise + float inaccuracy doesn't crash the app) |
| noshuffle | -ns or noshuffle |
When present, data won't be shuffled and will go in alphabetic order to their respective new folders. |
| copy | -c or copy |
When present, make copies instead of moving. Will generate a _split_ folder and generate the split inside it, preserving original directory structure and files. (May be slow and take a lot of space) |
Built under the MIT License.