Draft Bark finetuning

[erogol]

• Fine-tuning the semantic encoder.

from trainer import Trainer, TrainerArgs

dataset_config = BaseDatasetConfig(
    formatter="ljspeech", meta_file_train="metadata.csv", path="path/to/ljspeech/"
)

train_samples, eval_samples = load_tts_samples(
    dataset_config,
    eval_split=True,
    eval_split_max_size=4,
    eval_split_size=4,
)

trainer = Trainer(
    model=bark,
    config=bark_config,
    output_path="./",
    args=TrainerArgs(),
    train_samples=train_samples,
    eval_samples=eval_samples,
)
trainer.fit()

• Finetuning coarse and fine encoders
• Implement LoRA

[stale]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions. You might also look our discussion channels.

[erogol]

WIP

[stale]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions. You might also look our discussion channels.