Dataset → Fine-Tuning JSONL
Convert any CSV dataset to the strict JSONL format required by OpenAI, Anthropic, or Gemini fine-tuning. All processing happens in your browser.
Step 1: Input your CSV data
Step 3: Choose format
Upload a CSV and map columns to see output
Frequently asked questions
Which fine-tuning formats does the converter support?+
It outputs four JSONL formats: OpenAI fine-tuning (chat messages with optional system role), Anthropic prompt-completion pairs, Gemini fine-tuning (text_input/output), and a generic chat-completions format. You pick the target format and the tool builds one JSON object per line.
Does my dataset leave my browser?+
No. Parsing, column mapping, and conversion all run entirely in your browser with JavaScript — the CSV file is never uploaded to a server. This makes the tool safe to use with proprietary or sensitive training data.
What structure does my CSV need?+
A header row plus at least two columns: one for the input (the user prompt) and one for the expected output (the model response). You map the columns yourself after loading the file. For the OpenAI format you can optionally map a third column as the system message. Quoted fields containing commas are handled.
Why does the output have fewer lines than my CSV has rows?+
Rows where the mapped input or output column is empty are skipped, because every training example needs both a prompt and a completion. The tool shows how many rows were converted and how many were skipped so you can spot gaps in your data.
Is there a file size limit?+
The tool does not impose a hard limit, but because everything runs in your browser, very large files are constrained by your device’s memory. For typical fine-tuning datasets (thousands to tens of thousands of rows) it works fine; for very large corpora, consider splitting the file.
Is this converter free?+
Yes. It is completely free, requires no account, and you can copy the JSONL output or download it as a file ready to upload to your fine-tuning provider.