autonomous-learning-library
autonomous-learning-library copied to clipboard
cpnota
A PyTorch library for building deep reinforcement learning agents.
After #221 the way the presets are created has changed. This means that it is not possible to write anymore: ``` from all.presets import atari agents = [ atari.dqn(), atari.ddqn(),...
HI, I recently ran 'Hopper-v3' from mujoco_py (license is free to obtain now) in autonomous-learning-library by adopting the default setting from pybullet 'HopperBulletEnv-v0' and plotted the results as follows: (SAC...
When I run the following code: ``` from all.environments import AtariEnvironment from all.experiments import SingleEnvExperiment from all.presets import atari env = AtariEnvironment("Pong",device="cuda") preset = atari.ppo.env(env).device("cuda").hyperparameters().build() experiment = SingleEnvExperiment(preset, env) experiment.train(frames=2e6)...
There is a try-catch block here that emits a warning when using the multi-agent preset. The warning is created because creating a pytorch tensor out of a list of strings...
MultiagentPettingZooEnv is an odd and redundant name. This should probably just be MultiagentEnv.
Right not run_experiment doesn't accept multi-agent environments. Some solution to this should be created.
It makes using the continuous models annoying and makes it different than the actual paper, which isn't ideal
The documentation just kind of cuts off mid sentence? ``` class Aggregation(nn.Module): """ Aggregation layer for the Dueling architecture. https://arxiv.org/abs/1511.06581 This layer computes a Q function by combining an estimate...