Towards Soft-Prompt Tuning with Large Language Models | Vector Applied Intern Talks

Saeed’s talk examines a recent method for prompt engineering with large language models. During the internship, Saeed provided a reference implementation and experimentation framework based on PyTorch and HuggingFace for soft-prompt tuning with T5-large language models.
The implementation serves as a prototype or proof-of-concept for future development toward the large-scale application of soft-prompt tuning in language models. The preliminary experiment conducted on sentiment analysis tasks was promising.

Link from Video

Git Hub

[AzinAsgarian-kfcn]

Great summary and work, thanks Saeed!

[joshuasansom_software]

Thanks for the summary. Helped me understand things little better.