Node packVerificadoExterno

ComfyUI_NNT

NNT Neural Network Toolkit Custom Nodes for ComfyUI

Fuentegithub.com/inventorado/ComfyUI_NNT

Responsable@inventoradoUnratedcalidad

Perfil del paquete

Estado: Verificado
Publicador: @inventorado
Fuente: GitHub
Estrellas: 69

Léeme

Neural Network Toolkit (NNT) for ComfyUI

The Neural Network Toolkit (NNT) is an extensive set of custom ComfyUI nodes for designing, training, and fine-tuning neural networks. This toolkit allows defining models, layers, training workflows, transformers, and tensor operations in a visual manner using nodes.

This toolkit is especially useful in educational environments for learning and explaining neural network principles and architecture without code. Even people with no programming experience can use it to experiment with neural networks. I created this project as a practical learning tool while studying neural networks.

NNT allows you to quickly create and train various models, experiment with their structure, and observe the effect of changing their parameters. While it's a powerful learning and prototyping tool, please note it's a work in progress and not meant to replace PyTorch coding for production environments.

License

This project is licensed under the GNU General Public License v3.0 - see LICENSE for details.

Key Features

Visual node-based neural network design
Support for various layer types (Dense, Conv, RNN, Transformer)
Interactive training and fine-tuning capabilities
Real-time visualization of model architecture and training metrics
Tensor manipulation and analysis tools
Educational focus with immediate visual feedback
No coding required for basic operations

Installation

Prerequisites

ComfyUI installed and working
Python 3.8+
CUDA-compatible GPU (recommended)
PyTorch 2.0+
Basic understanding of neural networks

Installation can be done via ConfyUI Manager or manualy

cd ComfyUI/custom_nodes
git clone https://github.com/inventorado/ComfyUI_NNT.git
cd ComfyUI_NNT
pip install -r requirements.txt

On standalone version you can install requirements for embeded python using

cd ComfyUI_NNT
..\..\..\python_embeded\python.exe -m pip install -r requirements.txt

Example Workflows

Example workflows can be found in workflows directory under this custom node. They use ComfyUI-Jjk-Nodes for displaying text, you can install this by ConfyUI Manager - Install Missing Custom Nodes.

Linear Regression Example

A simple linear regression model demonstrating basic neural network construction and training. Shows how to create input layers, add linear layers, compile the model, and visualize training results.

CIFAR10 Convolutional Network

Advanced convolutional neural network for CIFAR10 image classification. Features data loading, model construction with conv layers, batch normalization, and pooling, followed by training and evaluation.

MNIST Digit Recognition

Classic MNIST handwritten digit classification using convolutional neural network. Demonstrates proper model architecture for digit recognition with convolutional layers, pooling, and dense layers.

MNIST Model Loading and Inference

Example of loading a pre-trained MNIST model and performing inference. Shows how to load model weights, process input data, and visualize prediction results.

Fashion-MNIST Autoencoder with RNN

Recurrent neural network (RNN) based autoencoder for Fashion-MNIST dataset. Demonstrates sequence processing capabilities with RNN layers for image reconstruction tasks.