Parameter-Efficient Fine-Tuning (PEFT)

PEFT is a popular technique used to efficiently finetune large language models for use in various downstream tasks. When finetuning with PEFT, the base model weights are frozen, and a few trainable adapter modules are injected into the model, resulting in a very small number (<< 1%) of trainble weights. With carefully chosen adapter modules and injection points, PEFT achieves comparable performance to full finetuning at a fraction of the computational and storage costs.

NeMo supports four PEFT methods which can be used with various transformer-based models. Here is a collection of conversion scripts that convert popular models from HF format to nemo format.

Learn more about PEFT in NeMo with the Developer Quick Start which provides an overview on how PEFT works in NeMo. Read about the supported PEFT methods here. For a practical example, take a look at the Step-by-step Guide.

The API guide can be found here

• Quick Start Guide
• Supported PEFT methods