chandar-lab/RLHive

Issues

• Update compatibility with gym environment

  #344 opened a year ago by delara38
  2

• Updating Notebooks with the new runner api

  #338 opened 2 years ago by hnekoeiq
  0

• Masking the DRQN

  #325 opened 2 years ago by mrsamsami
  0

• Visualization: having a cmap_name in plot_results kwargs is restrictive.

  #311 opened 2 years ago by JakeColor
  3

• pytorch version update ~=1.6 -> >=1.7

  #273 opened 2 years ago by sriyash421
  5

• `_max_steps_per_episode` in the main loop has the issue of not making `done=True`

  #285 opened 2 years ago by alirahkay
  2

• Reset the episode start inside act function or update function

  #297 opened 2 years ago by hnekoeiq
  1

• Change the test name

  #313 opened 2 years ago by hnekoeiq
  0

• Visualization: trying to overwrite previous images gives misleading error

  #310 opened 2 years ago by JakeColor
  0

• Bug with testing episodes

  #246 opened 2 years ago by hnekoeiq
  2

• Add DDPG, TD3 Agents

  #254 opened 2 years ago by dapatil211
  1

• CI/CD dependecy installation

  #283 opened 2 years ago by sriyash421
  1

• Add RLiable metrics

  #255 opened 3 years ago by dapatil211
  0

• Add the option to evaluate the agent before the training starts

  #277 opened 2 years ago by alirahkay
  0

• Buffer for on-policy algorithms like PPO, TRPO

  #271 opened 3 years ago by sriyash421
  1

• Add PPO Agent

  #253 opened 3 years ago by dapatil211
  0

• Jax DQN

  #252 opened 3 years ago by dapatil211
  0

• Add Atari 100k

  #251 opened 3 years ago by dapatil211
  0

• Recurrent DQN Agent

  #250 opened 3 years ago by dapatil211
  0

• Create a Recurrent Replay Buffer

  #249 opened 3 years ago by dapatil211
  0

• [README] Have a section in README on how to change config from CL. (Difference between complex objects and basic parameters)

  #165 opened 3 years ago by dapatil211
  0

• [README] Explain seeding and its impact on the speed

  #168 opened 3 years ago by dapatil211
  0

• Minatar doesn't run on cluster because of matplotlib/Tkinter issue

  #130 opened 3 years ago by dapatil211
  0

• Base.py #124 check if the train_mode is already True.

  #164 opened 3 years ago by dapatil211
  0

• use "cpu" if gpu is not available but device:cuda in config

  #107 opened 3 years ago by saikrishna-1996
  2

• Update Atari Env to v5

  #172 opened 3 years ago by dapatil211
  0

• Add agent-specific seed for multiple agents

  #167 opened 3 years ago by dapatil211
  0

• Add the global seed

  #158 opened 3 years ago by alirahkay
  0

• Do testing over a fixed number of testing episodes. (10 episodes per one test)

  #163 opened 3 years ago by dapatil211
  0

• Change qnet() name to something like representation_net()?

  #170 opened 3 years ago by dapatil211
  0

• Remove casting in buffer insert in legal_moves_rainbow

  #188 opened 3 years ago by dapatil211
  0

• add a small number to avoid numericall issues

  #181 opened 3 years ago by hnekoeiq
  0

• Change the initial replay buffer to the efficient replay buffer. Also change name of efficient buffer

  #166 opened 3 years ago by dapatil211
  0

• Dqn.py move update period schedule above target net update schedule

  #169 opened 3 years ago by dapatil211
  0

• Can't tuple an int

  #173 opened 3 years ago by saikrishna-1996
  0

• Use observation_space in pettingzoo instead of observation_spaces

  #187 opened 3 years ago by dapatil211
  0

• Dqn.py #202 Add a comment on breaking argmax tie.

  #171 opened 3 years ago by dapatil211
  0

• Testing epsilon should not be zero.

  #96 opened 3 years ago by alirahkay
  1

• Create an option for saving a run based on the number of training frames. (In the saving schedule)

  #133 opened 3 years ago by alirahkay
  1

• Fix logging multi-agent wandb params

  #137 opened 3 years ago by hnekoeiq
  0

• Change wandb settings to fork

  #131 opened 3 years ago by dapatil211
  0

• Add update period to the rainbow

  #122 opened 3 years ago by alirahkay
  0

• test episodes should not add to training steps

  #125 opened 3 years ago by hnekoeiq
  0

• log average testing performance

  #112 opened 3 years ago by hnekoeiq
  0

• Frame Stacking is still not working in `conv.py`

  #114 opened 3 years ago by alirahkay
  3

• One config should produce roughly the same results, fixing the seeds

  #95 opened 3 years ago by alirahkay
  0

• Debugging metrics

  #100 opened 3 years ago by dapatil211
  0

• Probe environments

  #99 opened 3 years ago by dapatil211
  0

• Create Debugging tools

  #98 opened 3 years ago by dapatil211
  0

• `EfficientCircularBuffer` uses *capacity*, whereas `CircularReplayBuffer` uses *size*.

  #97 opened 3 years ago by alirahkay
  1