added print_every for atari env score reporting

changed avg reward calculation

added monitor wrapper for atari envs

add action space seeding, as per sb3

add params to correctly calculate conv dims

code formatting

add kwargs to stackframes reset call

atari dependencies no longer required

fix dtype issue in converting arrays to tensors

compatibility with new learner interface

remove dependency on miniworld

improve examples

performance improvements

performance improvements

add value estimation function

better score handling

change actor/network architecture

remove adv norm and add swap_and_flatten func

change wrappers

change atari wrappers

return env.reset()

change batch sampling for ppo

add pixel environments

code cleanup

cnn factory added

ppo for atari

atari for ppo

pytorch observation wrapper

added new wrappers from sb3

better conv output dim calculation