finetune_mms_optimized.json

{
    "project_name": "mms_gujarati_finetuning",
    "push_to_hub": true,
    "hub_model_id": "Cossale/mms-tts-guj-train",
    "report_to": [
        "wandb"
    ],
    "overwrite_output_dir": true,
    "output_dir": "./tmp/vits_finetuned_guj",
    "dataset_name": "Cossale/gujarati-interspeech",
    "dataset_config_name": "default",
    "audio_column_name": "audio",
    "text_column_name": "text",
    "train_split_name": "train",
    "speaker_id_column_name": "speaker_id",
    "override_speaker_embeddings": true,
    "filter_on_speaker_id": 1,
    "full_generation_sample_text": "આજે દુબલીમાં ગરમી છે, શું આશ્ચર્ય છે",
    "max_duration_in_seconds": 20,
    "min_duration_in_seconds": 1.0,
    "max_tokens_length": 500,
    "model_name_or_path": "Cossale/mms-tts-guj-train",
    "preprocessing_num_workers": 40,
    "do_train": true,
    "num_train_epochs": 100,
    "gradient_accumulation_steps": 2,
    "gradient_checkpointing": true,
    "per_device_train_batch_size": 96,
    "learning_rate": 3e-5,
    "adam_beta1": 0.8,
    "adam_beta2": 0.99,
    "warmup_ratio": 0.03,
    "group_by_length": true,
    "do_step_schedule_per_epoch": true,
    "weight_disc": 3,
    "weight_fmaps": 1,
    "weight_gen": 1,
    "weight_kl": 1.5,
    "weight_duration": 1,
    "weight_mel": 35,
    "fp16": true,
    "seed": 456
}