config_se.json

{
    "model": "speaker_encoder",
    "run_name": "speaker_encoder",
    "run_description": "resnet speaker encoder trained with commonvoice all languages dev and train, Voxceleb 1 dev and Voxceleb 2 dev",
    "epochs": 100000,
    "batch_size": null,
    "eval_batch_size": null,
    "mixed_precision": false,
    "run_eval": true,
    "test_delay_epochs": 0,
    "print_eval": false,
    "print_step": 50,
    "tb_plot_step": 100,
    "tb_model_param_stats": false,
    "save_step": 1000,
    "checkpoint": true,
    "keep_all_best": false,
    "keep_after": 10000,
    "num_loader_workers": 8,
    "num_val_loader_workers": 0,
    "use_noise_augment": false,
    "output_path": "../checkpoints/speaker_encoder/language_balanced/normalized/angleproto-4-samples-by-speakers/",
    "distributed_backend": "nccl",
    "distributed_url": "tcp://localhost:54321",
    "audio": {
        "fft_size": 512,
        "win_length": 400,
        "hop_length": 160,
        "frame_shift_ms": null,
        "frame_length_ms": null,
        "stft_pad_mode": "reflect",
        "sample_rate": 16000,
        "resample": false,
        "preemphasis": 0.97,
        "ref_level_db": 20,
        "do_sound_norm": false,
        "do_trim_silence": false,
        "trim_db": 60,
        "power": 1.5,
        "griffin_lim_iters": 60,
        "num_mels": 64,
        "mel_fmin": 0.0,
        "mel_fmax": 8000.0,
        "spec_gain": 20,
        "signal_norm": false,
        "min_level_db": -100,
        "symmetric_norm": false,
        "max_norm": 4.0,
        "clip_norm": false,
        "stats_path": null
    },
    "datasets": [
        {
            "name": "voxceleb2",
            "path": "/workspace/scratch/ecasanova/datasets/VoxCeleb/vox2_dev_aac/",
            "meta_file_train": null,
            "ununsed_speakers": null,
            "meta_file_val": null,
            "meta_file_attn_mask": "",
            "language": "voxceleb"
        }
    ],
    "model_params": {
        "model_name": "resnet",
        "input_dim": 64,
        "use_torch_spec": true,
        "log_input": true,
        "proj_dim": 512
    },
    "audio_augmentation": {
        "p": 0.5,
        "rir": {
            "rir_path": "/workspace/store/ecasanova/ComParE/RIRS_NOISES/simulated_rirs/",
            "conv_mode": "full"
        },
        "additive": {
            "sounds_path": "/workspace/store/ecasanova/ComParE/musan/",
            "speech": {
                "min_snr_in_db": 13,
                "max_snr_in_db": 20,
                "min_num_noises": 1,
                "max_num_noises": 1
            },
            "noise": {
                "min_snr_in_db": 0,
                "max_snr_in_db": 15,
                "min_num_noises": 1,
                "max_num_noises": 1
            },
            "music": {
                "min_snr_in_db": 5,
                "max_snr_in_db": 15,
                "min_num_noises": 1,
                "max_num_noises": 1
            }
        },
        "gaussian": {
            "p": 0.0,
            "min_amplitude": 0.0,
            "max_amplitude": 1e-05
        }
    },
    "storage": {
        "sample_from_storage_p": 0.5,
        "storage_size": 40
    },
    "max_train_step": 1000000,
    "loss": "angleproto",
    "grad_clip": 3.0,
    "lr": 0.0001,
    "lr_decay": false,
    "warmup_steps": 4000,
    "wd": 1e-06,
    "steps_plot_stats": 100,
    "num_speakers_in_batch": 100,
    "num_utters_per_speaker": 4,
    "skip_speakers": true,
    "voice_len": 2.0
}