librispeech_tts_r_100.json

{
    "model_name_or_path": "/raid/yoach/tmp/artefacts/small-stable-speech-untrained/",
    "feature_extractor_name":"facebook/encodec_24khz",
    "description_tokenizer_name":"google-t5/t5-small",
    "prompt_tokenizer_name":"google-t5/t5-small",

    "push_to_hub": true,
    "hub_model_id": "ylacombe/stable-speech-mini",
    "report_to": ["wandb"],
    "overwrite_output_dir": true,
    "output_dir": "/raid/yoach/tmp/artefacts/training-mini/",


    "train_dataset_name": "blabble-io/libritts_r",
    "train_metadata_dataset_name": "stable-speech/libritts-r-tags-and-text-generated",
    "train_dataset_config_name": "clean",
    "train_split_name": "train.clean.100",

    "eval_dataset_name": "blabble-io/libritts_r",
    "eval_metadata_dataset_name": "stable-speech/libritts-r-tags-and-text-generated",
    "eval_dataset_config_name": "clean",
    "eval_split_name": "test.clean",


    "target_audio_column_name": "audio", 
    "description_column_name": "text_description",
    "prompt_column_name": "text",

    "max_eval_samples": 24,
    
    "max_duration_in_seconds": 29,
    "min_duration_in_seconds": 2.0,

    "add_audio_samples_to_wandb": true,
    "id_column_name": "id",

    "preprocessing_num_workers": 1,


    "pad_token_id": 1024,
    "decoder_start_token_id": 1025,


    "do_train": true,
    "num_train_epochs": 15,
    "gradient_accumulation_steps": 1,
    "gradient_checkpointing": true,
    "per_device_train_batch_size": 40,
    "learning_rate": 1e-4,
    "adam_beta1": 0.9,
    "adam_beta2": 0.999,
    "weight_decay": 0.1,

    "lr_scheduler_type": "cosine",
    "warmup_ratio":  0.1,


    "logging_steps": 102,
    "freeze_text_encoder": false,


    "do_eval": true, 
    "predict_with_generate": true,
    "include_inputs_for_metrics": true,
    "evaluation_strategy": "steps",
    "eval_steps": 3000,
    "save_steps": 3000,

    "per_device_eval_batch_size": 8,
    "generation_max_length": 400,

    "audio_encode_per_device_eval_batch_size":32,
    "dtype": "float16",
    "seed": 456,

    "dataloader_num_workers":16
}