Installation Guide¶
This guide covers the current ApexRL installation path.
Requirements¶
Python >= 3.10
PyTorch >= 2.0
Gymnasium >= 0.29
NumPy >= 1.24
TensorDict >= 0.6
From Source¶
git clone https://github.com/Atticlmr/Apex_rl.git
cd Apex_rl
pip install -e .
Using uv¶
git clone https://github.com/Atticlmr/Apex_rl.git
cd Apex_rl
uv pip install -e .
Optional Logging Extras¶
Install optional SDKs when using hosted logging backends:
pip install -e ".[wandb]"
pip install -e ".[swanlab]"
With uv:
uv pip install -e ".[wandb]"
uv pip install -e ".[swanlab]"
Verification¶
import apexrl
import torch
print(apexrl.hello())
print(torch.__version__)
Recommended Runtime Notes¶
TensorBoard logging works out of the box through the
tensorboarddependency.wandbandswanlabare optional extras because not every installation needs hosted experiment tracking.Structured observation support relies on
tensordictand is enabled in the default install.For local development, use a dedicated virtual environment to avoid PyTorch dependency conflicts.