Takuma Seno

@ChenDRAG Hi, thanks for the issue. `AsyncBatchEnv` will be removed from the next release (it's already removed from the master branch) because the asynchronous distributed training is planned to be...

Hmmm, I didn't expect there are many people using that feature. Let me revert them back and let me know if that works for you. Please note that `AsyncBatchEnv` is...

Could you show the pointer to `SB3MultipleInstanceEnv`? I guess it'd be in stable baselines 3, but I couldn't find it there.

@MishraIN Hi, sorry for the late response. `MOPO` is an algorithm class and `dynamics_reward_prediction_error_scorer` and `dynamics_prediction_variance_scorer` are the metrics for the dynamics models. You can use those metrics only for...

Do you mean this? https://github.com/takuseno/d3rlpy/blob/master/reproductions/offline/mopo.py

It's working on my end. ``` $ python reproductions/offline/mopo.py --dataset pendulum-random ``` Can you make sure you're using the latest script?

@jdchang1 Hello, thank you for the issue. Currently, I did not spend much time on checking MOPO's performance for now. But, very recently, the d4rl dataset conversion was fixed. https://github.com/takuseno/d3rlpy/commit/8e141c043db7a551875791c2c76db89cc140038f...

@TakuyaHiraoka Thanks for the info! I never realized they did a very tricky hack. It seems not very practical in general. I would rather add a classifier trained to estimate...

@Shentao-YANG Thanks for making this issue! Assuming that you're mentioning `d4rl-pybullet`, for the first question, d3rlpy currently does not have benchmark results. But, I believe we should have it since...

@nileshop22 Thank you for the issue. Technically it should be easy and already support the vision dataset. However, it seems the dataset is not available yet. For example, this URL...