README.md

Actuator Neural Network Training

This project aims to train neural network models for actuator control using PyTorch. It includes scripts for data preparation, model setup, training, validation, and evaluation.

Project Structure

This project is organized into several folders and files:

• datapreparation/: Contains the script which defines a custom dataset functions for handling data operations.
• model/: Contains the model. pth weights for each network.
• training.py for training and evaluation functions, and utilities for visualizing model predictions.
• actuator_net1.py: Contains an NN class described in paper "Learning agile and dynamic motor skills for legged robots"
• actuator_net2.py: Custom model.
• model_setup.py; Sets up models, optimizers, and learning rate schedulers for training.
• Validation.py: Contains functions for validating the models.
• metrics.py; Contains functions for computing evaluation metrics.
• run.py: The main script that ties everything together; it's used to execute the model's training and evaluation pipeline.
• inference and evaluate.py: Contains function that can assess model on unseen data.

Installation

To set up your environment for running this project, follow these steps:

1. Clone the repository:

   git clone https://github.com/HARISKHAN-1729/actuatornetwork.git
   cd actuatornetwork
   

2. Create and activate a virtual environment (optional but recommended):

    python -m venv venv
    source venv/bin/activate  # On Windows use `venv\Scripts\activate`
   

3. Install the required dependencies:

    pip install -r requirements.txt
   

4. Usage:

    python run.py
    python inference_and_evaluate.py