Last updated
Last updated
Clone the repository:
Install the dependencies:
This runs a prepared bash script with the necessary setup steps, with Python dependencies managed through Conda. Refer to the page if you encounter issues.
Readers of the bookFoundations of Deep Reinforcement Learning: please see .
SLM Lab uses PyTorch version 1.3.1 by default. For newer GPU cards, the versions of PyTorch and CUDA that come with SLM Lab default setup above may not be supported. In this case, you may bring-your-own-PyTorch by using to setup dependencies instead. This installs the same Python modules except for PyTorch and cudatoolkit. Use the following commands:
Non-image based environments can run on a laptop. Only image based environments such as the Atari games benefit from a GPU speedup. For these, we recommend 1 GPU and at least 4 CPUs. This can run a single Atari Trial
consisting of 4 Sessions
.
For desktop, a reference spec is GTX 1080 GPU, 4 CPUs above 3.0 GHz, and 32 GB RAM.
Thanks to for help with testing this on RTX A6000 and RTX A5000 GPUs.
The best way to run SLM Lab on Windows is to use a Bash shell/Linux subsystem. Credit to and for providing a detailed instruction PDF for doing so on Windows:
Although we do not recommend running SLM Lab on Google Colab or Jupyter notebooks (notebooks come with inherent limitations, e.g. no rendering/multi-processing), we have prepared an example notebook for illustration. Credit to for helping with this:
For details on how it works, refer to .
Another user has figured out installation in Colab in 2024 - an especially helpful update given the dependencies/hardware architecture changes in the past years. Please find their detailed instructions with additional examples here:
For cloud computing, start with an affordable instance of with a K80 GPU and 4 CPUs. Use the Deep Learning AMI with Conda when .