Your files stay on your device
Fine-tuning JSONL formatter and validator
Paste training data, validate it against the target's schema, and format it to one clean record per line. Every deviation named, nothing uploaded.
target-specific validation · cap 8 MB · no upload, the file never leaves this tab
Paste data or open a file, then press Validate & format.
Format and validate the data first: the split runs on the valid lines.
- openai-chat: each line one object with a non-empty "messages" array; roles limited to system, developer, user, assistant, tool; content a string or an array of text parts; optional numeric weight.
- openai-completions: each line one object with string (or string-array) "prompt" and "completion"; optional "suffix" and weight.
- Duplicate keys keep JSON.parse semantics, last value wins. And the loss ledger says so. Number spellings normalize (1.0 → 1) and that is reported too, with counts and the first line it happened on.
- Unknown top-level keys are preserved verbatim, named in the ledger: the formatter never drops your data, and never invents fields a target expects.
Two valid records; one with a training weight.
Line 2 has an unknown role; line 3 is not JSON; both are named, not dropped silently.
Prompt/completion pairs: the legacy fine-tuning shape.
A pretty JSON array becomes one record per line; the wrapper removal is reported.
Paste, drop, or type to begin. Everything stays on this device.