Trixy7/trixy_core/config/datasets/trainer.py

# -*- coding: utf-8 -*-
"""
Trainer-Konfiguration für ML-Modelle.
"""

from dataclasses import dataclass, field


@dataclass
class WakewordTrainerConfig:
    """Konfiguration für Wakeword-Training."""
    model_name: str = "custom"
    epochs: int = 100
    batch_size: int = 32
    learning_rate: float = 0.001
    validation_split: float = 0.2
    augmentation_enabled: bool = True
    early_stopping_patience: int = 10


@dataclass
class VoiceRecognitionTrainerConfig:
    """Konfiguration für Sprechererkennungs-Training."""
    model_name: str = "default"
    epochs: int = 100
    batch_size: int = 16
    learning_rate: float = 0.0001
    validation_split: float = 0.2
    embedding_size: int = 256
    augmentation_enabled: bool = True
    early_stopping_patience: int = 10


@dataclass
class AudioProcessingConfig:
    """Audio-Verarbeitungs-Konfiguration für Training."""
    sample_rate: int = 16000
    n_mels: int = 80
    n_fft: int = 1024
    hop_length: int = 256
    win_length: int = 1024
    fmin: float = 0.0
    fmax: float = 8000.0


@dataclass
class DataConfig:
    """Daten-Konfiguration für Training."""
    raw_data_directory: str = "trainer/data"
    processed_data_directory: str = "trainer/processed"
    cache_directory: str = "trainer/cache"
    min_samples_per_class: int = 10


@dataclass
class OutputConfig:
    """Ausgabe-Konfiguration für Training."""
    models_directory: str = "models"
    export_onnx: bool = True
    export_torchscript: bool = True
    model_password: str = ""  # Passwort für .pth Verschlüsselung


@dataclass
class TrainerConfig:
    """Haupt-Trainer-Konfiguration."""
    wakeword: WakewordTrainerConfig = field(default_factory=WakewordTrainerConfig)
    voice_recognition: VoiceRecognitionTrainerConfig = field(default_factory=VoiceRecognitionTrainerConfig)
    audio_processing: AudioProcessingConfig = field(default_factory=AudioProcessingConfig)
    data: DataConfig = field(default_factory=DataConfig)
    output: OutputConfig = field(default_factory=OutputConfig)
    device: str = "auto"  # auto, cpu, cuda, mps
    num_workers: int = 4
    seed: int = 42
    verbose: bool = True