gpt-lab

Benchmark and compare small GPTs for task-specific inference. Tests base, fine-tuned, and prompted models against shared eval datasets. Finds minimum viable model, compares fine-tuned vs prompted, and generates reports.