Configuration

Configuration#

The system supports flexible configuration for different deployment scenarios:

Local Development Configuration#

For local development, use configs/cpu.py which configures:

  • CPU Inference: PyTorch CPU-only and ONNX Runtime CPU configurations

Production Configuration#

For production deployment, use configs/cuda.py which supports:

  • GPU Acceleration: CUDA 12.1 support for high-performance inference

Configuration Components#

  • Feature Extractor: Wav2Vec2 model configuration and audio processing parameters

  • Inference Engine: ONNX Unitalker model paths and inference settings

  • Audio Splitting: Energy-based silence detection parameters

  • Postprocessing: Configurable pipelines with emotional profiles and effects

  • Server Settings: FastAPI server configuration, CORS, and worker settings

  • Logging: Comprehensive logging configuration with multiple output targets

Contents