downloading ambs

[adla700]

Hey, I have downloaded dreamerv3 previously in an environment. Can I just download jax and the other dependencies in the same environment or should I create some other environment? Hope you can help me with my confusion. Thanks!

[sacktock]

Hi, I think if you have downloaded dreamerv3 and got that running in another virtual environment, then you should be good to go - just use this environment. For the dopamine agents (IQN and rainbow) I would recommend setting up a separate virtual environment and install the dopamine dependencies. If you have any more questions then let me know.

…

On Tue, Sep 19, 2023 at 2:52 PM adla700 ***@***.***> wrote: Hey, I have downloaded dreamerv3 previously in an environment. Can I just download jax and the other dependencies in the same environment or should I create some other environment? Hope you can help me with my confusion. Thanks! — Reply to this email directly, view it on GitHub <#2>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AKTHW6GK3DR6BSNV46JAEKLX3GPTRANCNFSM6AAAAAA46LDBKE> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[sacktock]

This was the setup instructions I followed for Ubunto 22.04 LTS:
Install Conda for Ubuntu
https://linuxhint.com/install-anaconda-ubuntu-22-04/
conda create --name jax --clone base
Install JAX/CUDA on Ubuntu WSL 2
https://docs.nvidia.com/cuda/wsl-user-guide/index.html
https://catid.io/posts/jax/
https://docs.nvidia.com/deeplearning/cudnn/install-guide/index.html#troubleshoot
https://ubuntu.com/tutorials/enabling-gpu-acceleration-on-ubuntu-on-wsl2-with-the-nvidia-cuda-platform#5-enjoy-ubuntu-on-wsl
sudo apt-get install libfreeimage3 libfreeimage-dev
pip install gym[atari]==0.19.0
pip install gym[accept-rom-license]==0.19.0
pip install dm_control
pip install cloudpickle
pip install crafter
pip install numpy
pip install optax
pip install rich
pip install ruamel.yaml
pip install tensorflow-cpu
pip install tensorflow_probability
pip install chainer
pip install chainerrl
pip install Pillow
pip install cmake
pip install opencv-python

[adla700]

got it, thank you so much!

[adla700]

I noticed that some of the requirements have older versions that are not available, any alternative to be able to run the code?

[sacktock]

The requirements in requirements.txt are not strict, assuming the librariers that are out of date are backwards compatible you shiould be fine. However gym[atari]==0.19.0 is required for the code to work, you could use a newer version of gym but you would have to modify the source code where necessary.