Skip to main content

Module training_format

Module training_format 

Source
Expand description

Training data format generators.

Converts curriculum-ordered triples into JSONL and Alpaca instruction format for language model fine-tuning.

Structs§

TrainingExample
A single training example in JSONL format.

Functions§

section_counts
Count examples per section.
to_jsonl
Generate JSONL training data from a curriculum.
to_jsonl_random
Generate randomly-ordered JSONL from the same triples (baseline).