Batch inference provider for candidate generation models.

Evaluate batch responses with PandaLM

Collate examples for supervised fine-tuning.

Transforms a function that takes a callback into a lazy iterator (generator).