[OSPP Week-3] Improving RLDatasets.jl and implementing CQL algorithms.

[Mobius1D]

Work till now.

Thanks to the guidance of @findmyway I was able to build the following:

• Made a wrapper for d4rl, d4rl_atari and d4rl_pybullet offline datasets with sampling methods and install functions.
• Finalised many of the features that would be present in ReinforcementLearningDatasets.jl.
• Made a basic CQL implementation on top of SAC.

My work for the upcoming week would be:

• Finalize the CQL implementation on continuous cases and use RLDatasets d4rl dataset to implement a standard Experiment.

• Try reproducing the same with CQL Discrete on top of QRDQN.

• Enhance the functionality of the RLDatasets.jl

  • Expand dataset fetching feature to accomodate RL Unplugged and accomodate all datasets that are available.
  • Finalise and polish the package to improve its quality and usability.