Jiayi Weng
Jiayi Weng
You're right, and also reward shaping can somehow write in `gym.Wrapper` or `collector.preprocess_fn` (if treating F_D as a part of the environment). Currently we don't have any plan. But you...
Because in section `3.3 Policy Shaping` the authors doesn't claim that this algorithm is only compatible with DQN. My understanding is that this policy shaping method can be adapted to...
The first question is, yes, our group has some model-based works previously and we would like to refract and merge. I also have an RL work related to combining model-free...
> Does tianshou support distributed computation (e.g., on a cluster) for collecting experience? Currently we mainly support a single node of small-medium size of experiment, but you can use `RayVectorEnv`...
> Hi @Trinkle23897, > > I am running an experiment with Tianshou in a computer with multiple CPU cores and 2 GPUs. Is it possible to take advantage of that...
> By changing from VectorEnv to RayVectorEnv and using the specific arguments will Tianshou take care of the optimization? VectorEnv is not optimized. It uses the for-loop (typically for debugging)....
Wow, that's great! Sure, we will consider to provide some example scripts using Hydra. Thanks for your sharing!
Our code currently does not support this method. But we will consider your suggestion.
I replace torch-0.4 to torch-0.3.1 and import successfully. So this issue is pytorch0.4 's bug?