myelinio / moca

Continuous Meta-Learning without Tasks

This code accompanies the paper Continuous Meta-Learning without Tasks by James Harrison, Apoorva Sharma, Chelsea Finn, and Marco Pavone.

This repo also contains pytorch implementations of ALPaCA and PCOC (introduced in the above paper), simple Bayesian meta-learning algorithms for regression and classification respectively.

To install:

First, install the MOCA package via

pip install -e .

then install dependencies via

pip install -r requirements.txt

To train:

Simple sinusoid experiment:

python experiments/train.py --train.experiment_id=0 --train.seed=0 --train.experiment_name='example'

Note that GPU usage is disabled by default. To enable, add argument:

--data.cuda=1

where the argument corresponds to the device.

To test:

python experiments/test.py --model.model_name='7500.pt' --train.experiment_id=0 --train.train_experiment_name='example' --train.experiment_name='example'