Jiayi Weng comments

Results 303 comments of


                                            Jiayi Weng

How to use Tianshou with GPU and custom environment

It seems like all tensors in `batch.obs` are on `cuda:0` device. Did you customize the network part? Could you please paste a minimal working example to illustrate this issue? Thanks!

How to use Tianshou with GPU and custom environment

Close due to no response, feel free to re-open this issue.

Implement Decision Transformer for offline RL

> Decision Transformer is a recurrent/autoregressive-style Should we first fix tons of issues for RNN?

Implement Decision Transformer for offline RL

I can push gogoduan.

Implement Decision Transformer for offline RL

gogoduan says he will be available after 5.12 and through all summer break.

Implement Decision Transformer for offline RL

> I would be interested in this. Thanks in advance!

Implement Decision Transformer for offline RL

> I believe something is indeed wrong #486

Can we get the action probability in C51?

Feel free to inherit C51Policy: ```python class C51Visualize(C51Policy): def __init__(self, summary_writer, *args, **kwargs): super().__init__(*args, **kwargs) # native tensorboard SummaryWriter, not tianshou.utils.logger.TensorboardLogger # just to demonstrate how to customize logging process...

Can we get the action probability in C51?

My bad. ```diff - def forward(self, batch): - result = super().forward(batch) + def forward(self, batch, **kwargs): + result = super().forward(batch, **kwargs) ```

Can we get the action probability in C51?

Could you please paste your code here to reproduce the error?