THUDM/P-tuning
A method for parameter-efficient fine-tuning of large language models using trainable continuous prompt embeddings.
Not currently ranked — collecting fresh signals.
star history
P-tuning is a parameter-efficient fine-tuning technique for pre-trained language models. It introduces learnable continuous embeddings as trainable prompts while keeping most model parameters frozen. The approach enables effective adaptation of large language models like GPT to downstream tasks with fewer trainable parameters, supporting few-shot learning scenarios. The repository includes code and datasets for LAMA and SuperGLUE benchmarks.
Frequently asked
- What is THUDM/P-tuning?
- A method for parameter-efficient fine-tuning of large language models using trainable continuous prompt embeddings.
- Is P-tuning open source?
- Yes — THUDM/P-tuning is open source, released under the MIT license.
- What language is P-tuning written in?
- THUDM/P-tuning is primarily written in Python.
- How popular is P-tuning?
- THUDM/P-tuning has 939 stars on GitHub.
- Where can I find P-tuning?
- THUDM/P-tuning is on GitHub at https://github.com/THUDM/P-tuning.