training_config.json 1.5 KB

12345678910111213141516171819202122232425262728293031323334353637383940414243444546474849505152535455565758596061
  1. {
  2. "trainer_name": "voice_recognition_trainer",
  3. "model_name": "basic_voice",
  4. "output_dir": "./models/voice_recognition",
  5. "data_dir": "./trainer/data/voice_recognition",
  6. "config_dir": "./config",
  7. "batch_size": 32,
  8. "learning_rate": 0.001,
  9. "num_epochs": 200,
  10. "min_epochs": 20,
  11. "max_epochs": 1000,
  12. "early_stopping_patience": 20,
  13. "model_format": ".pth",
  14. "use_mixed_precision": false,
  15. "gradient_clip_norm": 5.0,
  16. "weight_decay": 0.0001,
  17. "validation_split": 0.2,
  18. "test_split": 0.1,
  19. "shuffle_data": true,
  20. "num_workers": 4,
  21. "pin_memory": true,
  22. "sample_rate": 16000,
  23. "bit_depth": 16,
  24. "audio_length": 3.0,
  25. "n_mels": 40,
  26. "n_fft": 512,
  27. "hop_length": 160,
  28. "win_length": 400,
  29. "use_augmentation": false,
  30. "noise_factor": 0.05,
  31. "speed_factor": 0.05,
  32. "pitch_factor": 0.02,
  33. "volume_factor": 0.1,
  34. "use_password_protection": true,
  35. "password": null,
  36. "encryption_algorithm": "AES256",
  37. "log_level": "INFO",
  38. "save_checkpoints": true,
  39. "checkpoint_interval": 10,
  40. "validate_interval": 1,
  41. "log_interval": 100,
  42. "use_lightweight_validation": true,
  43. "lightweight_sample_ratio": 0.3,
  44. "lightweight_max_samples": 1000,
  45. "comprehensive_validation_epochs": null,
  46. "device": "cuda",
  47. "use_distributed": false,
  48. "num_gpus": 1,
  49. "resume_from_checkpoint": false,
  50. "checkpoint_path": null,
  51. "freeze_backbone": false,
  52. "use_scheduler": true,
  53. "scheduler_type": "cosine",
  54. "custom_params": {
  55. "mode": "train",
  56. "debug": false,
  57. "model_format": ".pth",
  58. "embedding_dim": 192,
  59. "use_password_protection": false
  60. }
  61. }