Utils

This module provides a collection of utility scripts and helper classes designed to support various aspects of reinforcement learning experiments, including:

• Environment wrappers that add noise to actions and observations, and modify rewards.

• Custom activation functions to enhance neural network expressiveness.

• Tools for collecting, processing, and visualizing evaluation results across models and environments.

• Functions to create and configure Gym environments with consistent seeding and wrapping.

• Neural network utilities, including MLP and Bayesian MLP architectures and optimizer/loss creators.

• General-purpose tensor conversion and shape-checking utilities for PyTorch and NumPy interoperability.