@choihrM3 noticed an indexing issue when selecting the last-step fidelities/rewards from an episode. This indexing issue is caused because in the scripts alg_config.build() calls the step function of the environment one time, and thus adds to the environment's transition_history list and makes the previous logic for indexing off by one.
This PR uses episode_id to select the fidelity and reward from the last step of each episode for plotting. This should be more robust.
This PR also removes the try/except code for loading .npy files, which to the best of my knowledge are unused.
@choihrM3 noticed an indexing issue when selecting the last-step fidelities/rewards from an episode. This indexing issue is caused because in the scripts
alg_config.build()
calls thestep
function of the environment one time, and thus adds to the environment'stransition_history
list and makes the previous logic for indexing off by one.This PR uses
episode_id
to select the fidelity and reward from the last step of each episode for plotting. This should be more robust.This PR also removes the try/except code for loading .npy files, which to the best of my knowledge are unused.