Train, fine-tune, and evaluate neural graphics models with Neural Graphics Model Gym | Code-Along

Join us for a hands-on code-along exploring **Neural Graphics Model Gym**, an open-source toolkit for training, fine-tuning, evaluating, quantizing, and exporting neural graphics models. Using **Neural Frame Rate Upscaling (NFRU)** as our example, we’ll walk through the complete neural graphics development workflow, from a PyTorch model to a deployment-ready artifact for the Arm neural graphics stack. Follow along as we: * Train and fine-tune an NFRU model using PyTorch and Jupyter notebooks * Evaluate model quality and compare generated frames against ground truth * Explore training and evaluation metrics * Apply quantization-aware training for an INT8 deployment path * Export the model to **VGF** for use with ML Extensions for Vulkan * Inspect the exported model with **Model Explorer** * See how the workflow can be adapted to neural graphics models trained on your own game content Whether you’re working in machine learning, real-time graphics, or game development, this code-along shows how familiar ML tools can be used to start experimenting with neural graphics today. **Resources** Neural Graphics Model Gym Learning Path: https://learn.arm.com/learning-paths/mobile-graphics-and-gaming/model-training-gym-nfru/ Neural Graphics Model Gym: https://github.com/arm/neural-graphics-model-gym Model Gym examples and notebooks: https://github.com/arm/neural-graphics-model-gym-examples Explore Arm Neural Graphics: https://developer.arm.com/mobile-graphics-and-gaming/neural-graphics Events we mentioned: https://www.wearedevelopers.com/world-congress-north-america https://developer.arm.com/events/arm-create-shanghai https://developer.arm.com/events/arm-create-shenzhen https://armeverywhere.cn/en/ #Arm #NeuralGraphics #MachineLearning #PyTorch #Vulkan #GameDevelopment #GraphicsProgramming