MAML-TF2.2

• paper : Model-Agnostic Meta-Learning for Fast Adaptation of Deep Networks

Requirements

pip install requirements.txt

Datasets

Omniglot

1. Download Omniglot dataset and extract the contents of images_background.zip and images_evaluation.zip to ./datasets/omniglot/

git clone git@github.com:brendenlake/omniglot.git

datasets/omniglot
|-- images_background
|-- images_evaluation

2. Modify parameters in config.py, such as batchsize, lr ,update step, ...

• train model -> please set cfg.mode = True
• test model -> please set cfg.mode = False

3. Run the main python script(trian or test)

python main.py