CP-Prompt: Composition-Based Cross-modal Prompting for Domain-Incremental Continual Learning

This repository is our PyTorch implementation of CP-Prompt.

Requirements

pip install -r requirements.txt

How to run

You can run the CP-Prompt with the following commands:

# CDDB
python main.py --config configs/prefix_one_prompt/cddb.json >> logs/prefix_one_prompt/cddb.log 2>&1
# CORe50
python main.py --config configs/prefix_one_prompt/core50.json >> logs/prefix_one_prompt/core50.log 2>&1
# Domainnet
python main.py --config configs/prefix_one_prompt/domainnet.json >> logs/prefix_one_prompt/domainnet.log 2>&1

Acknowledgement

Our source code and data processing are built heavily based on the code of Sprompt ([https://github.com/iamwangyabin/S-Prompts]).

The data set download address is provided in the paper.