Skip to content

fyntra702/neuralforge

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

2 Commits
 
 

Repository files navigation

🧠 NeuralForge — Deep Learning Framework for Unity

Production-ready framework for neural network inference, training, and deployment in Unity. Built for AI-driven games and realtime intelligent systems.

✨ Features Neural Networks: Dense, convolutional, recurrent layers with multiple activations

GPU Acceleration: Compute shader backend for realtime performance

Model Import: ONNX, TensorFlow, PyTorch support

Training Pipeline: In-editor training with visual monitoring

Unity Integration: Component-based, sensor inputs, AI controllers

Visualization: Tensor heatmaps, training dashboards, network inspection

📦 Requirements Unity 2022.3 LTS+

Compute Shader support

4GB+ VRAM recommended

🛠 Installation Package Manager → Add from Git URL:

text https://github.com/yourusername/neural-forge.git 🚀 Quick Start csharp // Create network var network = gameObject.AddComponent(); network.Architecture = new Layer[] { new DenseLayer(128, Activation.ReLU), new DenseLayer(64, Activation.ReLU), new DenseLayer(10, Activation.Softmax) };

// Train var trainer = network.GetTrainer(); yield return trainer.TrainAsync(trainingData, epochs: 100);

// Predict Tensor output = network.Predict(input); int predictedClass = output.ArgMax(); 🧩 Components Core: NeuralNetwork, TensorProcessor, ModelImporter Sensors: VisionSensor, AudioFeatureExtractor, LidarPointCloud AI: PolicyAgent, BehaviorPredictor, UtilityNetwork

🎮 Examples Smart NPC Behavior:

csharp Tensor context = GatherEnvironmentContext(); Tensor actionScores = decisionMaker.Evaluate(context); AIAction bestAction = SelectAction(actionScores); Style Transfer:

csharp Tensor stylized = styleTransferModel.Predict(content); TensorUtils.TensorToTexture(stylized, outputArt); 🔧 Performance csharp // GPU acceleration NetworkConfig.EnableGPUInference = true;

// Model quantization var quantizedModel = network.Quantize(QuantizationType.Int8); 🧪 Samples Image Classification (MNIST)

Reinforcement Learning (CartPole)

Style Transfer

Smart NPCs

Procedural Generation

🛣 Roadmap Transformer architectures

Mobile deployment

Cloud training integration

Federated learning

📜 License

MIT © 2026 NeuralForge Contributors

About

No description or website provided.

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors