An Implementation of Model Agnostic Meta Learning. In particular, I've implemented the sinusoid experiment from the paper.
To setup:
pip install -e .
Please create an issue if you find any missing dependencies!
To train a network
python regression_example.py
Useful flags:
--device
: "cpu" or "gpu". Defaults to "cpu".
You can set other flags optionally. Check regression_example.py
to see the full list of flags.
To demo a network:
python regression_example.py --demo --results-dir <path to directory where network is stored>
The network is assumed to be named regressor.pt
. The training script will automatically save the network in results/regressor.pt
Results on low end of sinusoid task distribution
Results on high end of sinusoid task distribution