/MAML-Pytorch

Elegant PyTorch implementation of paper Model-Agnostic Meta-Learning (MAML)

Primary LanguagePython

MAML-Pytorch

PyTorch implementation of the supervised learning experiments from the paper: Model-Agnostic Meta-Learning (MAML): https://arxiv.org/abs/1703.03400

Both MiniImagenet and Omniglot Datasets are supported! Have Fun~

For Tensorflow Implementation, please visit HERE. For First-Order Implementation, Reptile namely, please visit HERE.

Platform

  • python: 3.x
  • Pytorch: 0.4+

MiniImagenet

Howto

  1. download MiniImagenet dataset from here, splitting: train/val/test.csv from here.
  2. extract it like:
miniimagenet/
├── images
	├── n0210891500001298.jpg  
	├── n0287152500001298.jpg 
	...
├── test.csv
├── val.csv
└── train.csv
  1. modify the path in miniimagenet_train.py:
		# batchsz here means total episode number
		mini = MiniImagenet('/hdd1/liangqu/datasets/miniimagenet/', mode='train', n_way=n_way, k_shot=k_shot, k_query=k_query,
		                    batchsz=10000, resize=imgsz)
		...
		mini_test = MiniImagenet('/hdd1/liangqu/datasets/miniimagenet/', mode='test', n_way=n_way, k_shot=k_shot, k_query=k_query,
				                    batchsz=600, resize=imgsz)

to your actual data path.

  1. just run python miniimagenet_main.py and running screenshot is as follows: screenshot-miniimagetnet

Benchmark

Model Fine Tune 5-way Acc. 20-way Acc.
1-shot 5-shot 1-shot 5-shot
Matching Nets N 43.56% 55.31% 17.31% 22.69%
Meta-LSTM 43.44% 60.60% 16.70% 26.06%
MAML Y 48.7% 63.11% 16.49% 19.29%
Ours Y 48.1% 62.2% - -

Ominiglot

Howto

run python omniglot_train.py, the program will download omniglot dataset automatically.

decrease the value of meta_batchsz to fit your GPU memory capacity.