Skip to content
Train / Validation / Test Dataset Splitter
Shuffle lines of CSV or JSONL data with a fixed seed and split them into train, validation and test sets by percentage. Download each file.
🔒 Processed locally in your browser — we don’t upload your input.
How to use the Dataset Splitter
Machine learning models must be evaluated on data they did not see during training, otherwise accuracy numbers are misleading. The standard practice is to split a dataset into a training set, a validation set used for tuning, and a held-out test set for the final score. This tool does that for any line-based data such as JSONL fine-tuning files, CSV rows or plain text samples.
Paste your data or open a file, set the percentages (for example 80/10/10), and choose a random seed. Lines are shuffled with a deterministic seeded generator, so the same seed always produces the same split, which keeps experiments reproducible and lets teammates recreate your exact sets. If your CSV has a header row, enable the header option and it will be copied to the top of each output instead of being shuffled into the data. Empty lines are dropped automatically.
Each split can be copied or downloaded as its own file with a sensible name. Counts are shown so you can confirm the proportions. Everything runs in the browser, so you can split private or licensed datasets without uploading them anywhere.
Frequently asked questions
What split ratio should I use?
80/10/10 or 70/15/15 are common. Small datasets may use cross-validation instead.
Why use a seed?
A seed makes the shuffle reproducible, so the same input and seed give identical splits.
Does it keep CSV headers?
Yes, enable the header option to copy the header to each output file.
Is it stratified?
No, it is a random split. For heavily imbalanced labels, check class balance afterwards.
About this tool
Shuffle lines of CSV or JSONL data with a fixed seed and split them into train, validation and test sets by percentage. Download each file. People also look for: train test split, dataset splitter online, validation split, split jsonl.
Related searches