metadrop-pytorch

Unofficial PyTorch Implementation of "Meta Dropout: Learning to Perturb Latent Features for Generalization" (ICLR 2020)

To Run: python main.py --phi

To Run MAML: python main.py

To get the data: python data_haebom/get_data.py

Results

	Omni. 1shot (main paper)	Omni. 1shot (Ours)
MAML	95.23	96.76
Meta-dropout	96.63	97.12

See the runs and report on the results.

To Install

conda install pytorch torchvision torchaudio cudatoolkit=10.2 -c pytorch

To Do

• develop the 'minimal' branch
• set a smaller runtime that can still track performance, so as to develop/debug faster