TorchRL

FluidGym environments are compatible with TorchRL, allowing seamless integration with various RL libraries that support TorchRL. This is especially useful to leverage PyTorch’s automatic differentiation capabilities for RL.

Due to the complexity of the TorchRL interface, we provide only a minimal example here from examples/interfaces/torchrl.py:

import fluidgym

from fluidgym.integration.torchrl import TorchRLFluidEnv

env = fluidgym.make("CylinderJet2D-easy-v0")
env.seed(42)

# For the TorchRL interface, wrap the FluidGym environment
trl_env = TorchRLFluidEnv(env)

# use with torchrl ...