A generalized framework for subspace tuning methods in parameter efficient fine-tuning.

[ECCV 2024] Soft Prompt Generation for Domain Generalization

Library for soft prompt tuning

Implementation of the following parameter-efficient fine-tuning methods on GPT-2 for summarization: Soft prompt tuning by optimizing prefix embeddings, LoRA, and fine-tuning only the last classifier layer while keeping the rest of the model frozen