Luca Anzalone

Hi, the error occurs when the spawn location is already occupied: it happens when some other actor (typically a vehicle) is placed on that location, and sometimes even when restarting...

Hi, I don't know exactly how much GPU mem is needed but, unfortunately, is plenty. What you can try is to edit [main.py](https://github.com/Luca96/carla-driving-rl-agent/blob/master/main.py). In particular you could set `timesteps=256` and/or...

@nguyenvantruong1905, check the following: - Do you set the environment variables before loading tensorflow, i.e. before `import tensorflow as tf`? - Do you have a CUDA capable GPU? If so,...

Hi, - The full training tooks around ten days, and it also depends on your hardware. - It's described in the results section of the Readme, but basically was a:...

Hi, without the full traceback is a little bit difficult to understand where the error is, could you provide it?

Hi, you can try a couple of things: - Change the loss function to `binary_crossentropy`. - And/or replace all the `keras` related imports with `tensorflow.keras`, instead. Let me know if...

Hi @SURABHI-GUPTA, unfortunately I have only developed the keras version for this project. Maybe you can try yourself to migrate the code to pytorch, and sent a pull-request. 😉

Hi @hmalpani , thanks for the help but increasing the task stack size still not works: I increased it to 16k but now I get a different error: ``` I...

Hello @david-cermak, thank you very much for your help! I finally tested the changes you suggested, and I can say that the code works without any crashes. In addition, I...

Hi @Andrea010725. You can use the [CARLACollectWrapper](https://github.com/Luca96/carla-driving-rl-agent/blob/master/rl/environments/carla/environment.py#L667) to record all the inputs, outputs, and intermediate values (e.g., rewards, returns, etc) which are all saved as traces (i.e., `npz` compressed files...