A PyTorch implementation of the paper "Probabilistic task modelling for meta-learning" published at International Conference on Uncertainty in Artificial Intelligence (UAI) 2021.
- PyTorch 1.0 or above with or without GPU support
- Tensorboard
Two popular datasets: Omniglot and mini-ImageNet are considered in the paper. Note that the image size on both datasets is set to 64-by-64 pixel2 to ease the design of VAE architecture.
The implementation also has Tensorboard integrated to monitor the training progress.