https://github.com/axect/pytorch_derivative_test

https://github.com/axect/pytorch_derivative_test

Last synced: 2 months ago
JSON representation

• Host: GitHub
• URL: https://github.com/axect/pytorch_derivative_test
• Owner: Axect
• License: mit
• Created: 2024-06-11T00:49:02.000Z (12 months ago)
• Default Branch: main
• Last Pushed: 2024-06-11T02:00:00.000Z (12 months ago)
• Last Synced: 2025-03-28T19:46:35.687Z (2 months ago)
• Language: Python
• Size: 1.65 MB
• Stars: 2
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# PyTorch Derivative Test

This project aims to verify whether derivatives are well-trained when training a neural network using PyTorch. By comparing the learned derivatives with the true derivatives, we can assess the effectiveness of the training process.

## Installation

To use this project, you can easily set up the same environment using [`uv`](https://github.com/astral-sh/uv):

```bash

# Create virtual env

uv venv

# Sync

uv pip sync requirements.txt

# Activate

source .venv/bin/activate

```

This command will install all the required dependencies specified in the `requirements.txt` file.

Alternatively, you can use `pip` to install the dependencies:

```bash

pip install -r requirements.txt

```

## Usage

To run the project, simply execute the `equation_net.py` script:

```bash

python equation_net.py

```

The script will generate random data, train the neural network, and plot the learned functions and derivatives along with the true functions and derivatives.

## Results

The project generates two plots:

1. Output and derivatives with respect to x (y=0):

![Output with y=0](output_x.png)

2. Output and derivatives with respect to y (x=0):

![Output with x=0](output_y.png)

The solid lines represent the learned functions and derivatives, while the dotted lines represent the true functions and derivatives. By comparing the solid and dotted lines, you can assess how well the neural network has learned the functions and their derivatives.

## License

This project is licensed under the MIT License. See the [LICENSE](LICENSE) file for more information.