For the code used for training the individual tasks and the meta-RL Agents please refer to:

https://github.com/rlworkgroup/garage

For the domains used please look at:

Ant:
https://github.com/cbfinn/maml_rl/blob/9c8e2ebd741cb0c7b8bf2d040c4caeeb8e06cc95/rllab/envs/mujoco/ant_env_rand.py

Cheetah:
https://github.com/rlworkgroup/garage/blob/93d1d6f0d546b544ab52bc399cacad3f0c696849/src/garage/envs/mujoco/half_cheetah_vel_env.py

KrazyWorld:
https://github.com/bstadie/krazyworld

CartPole:
https://github.com/openai/gym/blob/master/gym/envs/classic_control/cartpole.py

MiniGrid:

https://github.com/maximecb/gym-minigrid


For evaluation we made use of the meta evaluator provided in the "garage" library but introducing test tasks and number of gradient updates manually.

https://github.com/rlworkgroup/garage/blob/93d1d6f0d546b544ab52bc399cacad3f0c696849/src/garage/experiment/meta_evaluator.py

