amp-training

What problem does it solve?

Automates end-to-end training of neural amplifier models from paired DI and reamped guitar signals using PyTorch.

Core Features & Use Cases

• End-to-end workflow: capture, preprocessing, model training, evaluation, and export to RTNeural JSON or NAM formats.
• Data handling: supports 48 kHz / 24-bit mono captures, normalization to [-1,1], chunking, and train/val/test splits.
• Model architecture: LSTM-based networks with a lightweight head enabling practical inference on typical hardware.
• Deployment readiness: export options and integration with REVITHION STUDIO tools for real-time use.

Quick Start

Run the NAM workflow to capture input and target signals and execute the end-to-end training pipeline to produce a deployable neural-amp model.