This repository provides an extension to the pyMeta library, which can be used to run meta-learning experiments with the CORe50 continual learning dataset. In this example, training is done via Google Colab.
- Download the dataset: core50_imgs.npz & paths.pkl
- Download pyMeta
- Unzip and upload pyMeta to Google Drive
- Inside the folder 'pyMeta-master/datasets' create a new folder called 'core50'
- Upload the files core50_imgs.npz & paths.pkl to the folder 'pyMeta-master/datasets/core50'
- Download this repository
- Upload the folder 'core50' of this repository to 'pyMeta-master/pyMeta'
- Upload the file 'core50_metatrain.py' of this repository to the root folder 'pyMeta-master'
- Upload the file 'COReTrain.ipynb' to Google Drive and open the file in Google Colab
- In Google Colab make sure to change runtime type to GPU
Happy training!