genrl
genrl copied to clipboard
SforAiDl
A PyTorch reinforcement learning library for generalizable and reproducible algorithm implementations with an aim to improve accessibility in RL
HER Wrappers
Wrt #171 Have added a `HERTrainer`, `HERGoalEnvWrapper`, and a `HERWrapper` for the replay buffer. Some changes in the locations of the tests might be needed.. Wasnt too sure of where...
Stuff implemented: - Added BCQ under genrl/agents/offline - BCQ inherits from `OffPolicyAgentAC`. Architecture was very similar to TD3. Major differences were that the actor took in both state and action...
Wrt #375 Mades some changes in discount.py (Really silly mistakes) Now the 2 agents are training
There seem to be some vulnerabilities in our code that might fail easily. I suggest adding more unit tests for the following: - Custom agents (there's only VPG and PPO...
DQN is also not training but that'll be addressed after DQN is restructured.
There might be some shape related errors or we're missing something. Either that or hyperparameters need to be tuned.
We're moving from the current docstring style to the Google docstring style. Please refer to [DQN](https://github.com/SforAiDl/genrl/blob/master/genrl/deep/agents/dqn/base.py) and [this](https://sphinxcontrib-napoleon.readthedocs.io/en/latest/example_google.html) for an idea. This is a pretty long issue and pretty important...
Right now, we have about 50+ code smells on Code Climate. A lot of these are pretty hard to avoid. We should keep this as a long term issue. If...
Documentation on How can we extend agents? @sampreet-arthi added a great tutorial which can be referred #259