Table of Contents

Enum TrainingDatasetFormat

Namespace
LMKit.Finetuning
Assembly
LM-Kit.NET.dll

Dataset file format for TrainingDataset loaders.

public enum TrainingDatasetFormat

Fields

Auto = 0

Detect the format from the file extension and content.

ChatJsonl = 1

One JSON object per line: {"messages":[{"role","content"}]}. The OpenAI fine-tuning shape.

ShareGpt = 2

A JSON array of {"conversations":[{"from","value"}]}.

Alpaca = 3

A JSON array of {"instruction","input","output"}.

PlainText = 4

A plain-text corpus split into samples on a <SFT> delimiter, or on blank-line paragraph boundaries.

Share