config.json

{
    "name": "FastSpeech2",
    "n_gpu": 1,
    "arch": {
        "type": "FastSpeechModel",
        "args": {
            "vocab_size": 300,
            "max_seq_len": 3000,
            "encoder_dim": 256,
            "encoder_n_layer": 4,
            "encoder_head": 2,
            "encoder_filter_size": 1024,
            "decoder_dim": 256,
            "decoder_n_layer": 4,
            "decoder_head": 2,
            "decoder_filter_size": 1024
        }
    },
    "data": {
        "train": {
            "batch_size": 20,
            "batch_expand_size": 24,
            "num_workers": 5,
            "datasets": [
                {
                    "type": "LJDataset",
                    "args": {
                        "mel_ground_truth": "/kaggle/input/fastspeecch-dataset/data/data/mels",
                        "alignment_path": "/kaggle/input/fastspeecch-dataset/data/data/alignments",
                        "pitch_path": "/kaggle/input/fastspeecch-dataset/data/data/pitch",
                        "energy_path": "/kaggle/input/fastspeecch-dataset/data/data/energy",
                        "data_path": "/kaggle/input/fastspeecch-dataset/data/data/train.txt",
                        "text_cleaners": ["english_cleaners"],
                        "batch_expand_size": 24
                    }
                }
            ]
        }
    },
    "test_texts": ["My name is Kirill Korolev and i am going to smash your head with my precious hammer",
                   "English you speak it motherfucker", 
                   "pull up on the tank and e ya edu v boi" ],
    "optimizer": {
        "type": "AdamW",
        "args": {
            "lr": 3e-4,
            "weight_decay": 1e-6,
            "eps": 1e-7
        }
    },
    "loss": {
        "type": "FSLoss",
        "args": {}
    },
    "lr_scheduler": {
        "type": "OneCycleLR",
        "args": {
            "steps_per_epoch": 3000,
            "epochs": 48,
            "anneal_strategy": "cos",
            "max_lr": 3e-4,
            "pct_start": 0.1
        }
    },
    "trainer": {
        "batch_expand_size": 24,
        "epochs": 48,
        "save_dir": "saved/",
        "save_period": 5,
        "verbosity": 2,
        "monitor": "min loss",
        "early_stop": 50,
        "visualize": "wandb",
        "wandb_project": "fastspeech_project",
        "run_name": "MyanSpeech2_fast_epoch",
        "sample_rate": 22050,
        "len_epoch": 3000,
        "log_step": 50,
        "grad_norm_clip": 1,
        "batch_accum_steps": 1
    }
}