Kinal Mehta

After solving the previous issue by setting up the environment to be callable in the `run_dqn.py` file, the ExperimentConfig becomes non-serializable. So it is not possible to run the program...

Results on classical gym environments can be checked here. https://wandb.ai/kinalmehta/jax-cleanrl/reports/C51-JAX-vs-Pytorch-on-Classical-Gym-Environments--VmlldzoyNDQ3OTk5 We see a speed-up of about ~30% in the JAX version compared to Pytorch.

Hi @DavidSlayback Based on the repos you shared, it seems to be using very old `mujoco` versions, but I guess the environments should be available in the latest `gym` versions....

The issue seems to be because of the parameters being passed to _policy method which are not TF objects and @tf.function only seems to support TF objects based on following...

a similar issue occurred to me I believe the subfolder config is not being handled since 0.8.0

What you're saying is correct. Just adding the CSS support for colours should ideally work. But it seems the repo I linked is no longer maintained and is not present...

Hey. Just got to know that the highlight plugin will be uploaded to marketplace in few weeks. So maybe just adding proper color cas would be a better way to...

Hey @SlyBouhafs, The original author has still not pushed the highlight plugin on logseq-marketplace. Any suggestion from your side on how to proceed? Thanks

Hey @SlyBouhafs, I've got this working perfectly in my setup and have created pull request #14. Cheers Kinal

Here is the benchmark report on atari environments https://wandb.ai/kinalmehta/jax-cleanrl/reports/C51-JAX-vs-Pytorch-on-Atari-Environments--VmlldzoyNjkyNDY0 **Important observations:** - BeamRider performance is bad compared PyTorch version - Breakout performance almost matches PyTorch variant but is still a...