Skip to content

Latest commit

 

History

History
51 lines (33 loc) · 1.2 KB

installation.rst

File metadata and controls

51 lines (33 loc) · 1.2 KB
Raw

Installation

Prerequisites

  • FleetRL requires Python >=3.8

Note

Python >= 3.10 is strongly recommended.

Note

The creation of a virtual environment is strongly recommended. To be able to use GPU compute, CUDA drivers must be installed (11.8 was mostly used during development).

Installation via Github repository:
  • Unzip the package
  • Rename directory from FleetRL-master to FleetRL
  • cd into /FleetRL
  • pip install -r requirements.txt

Note

On remote environments on vast.ai it can be necessary to run pip install -U numpy prior to installing FleetRL

Miniconda Windows

In this example, FleetRL can be installed completely from scratch, only Miniconda is required. Run the commands below consecutively.

conda create -n **environment_name** python=3.10
conda activate **environment_name**
pip install jupyter
jupyter notebook

Inside the Jupyter Notebook, being in the FleetRL directory:

!pip install -r requirements.txt
# restart kernel
import FleetRL

At this point, the complete_pipeline.ipynb should run completely. To use GPU, CUDA must be properly configured.