About the dataset used in training policy generator

[eruka-w]

@click.command() @click.option('--train_path', help='Huggingface dataset name', required=True ) @click.option('--out_dir', default='../checkpoints/', help='Output directory',show_default=True) @click.option('--batch_size', default=8, help='Batch size',show_default=True) @click.option('--lr', default=2e-4, help='Learning rate',show_default=True) @click.option('--random_seed', default=1, help='Random seed',show_default=True) In the training code for the policy generator, it seems that a Hugging Face dataset is required, but the readme doesn't specify which one. Could you please inform me which dataset you are using, or if there's something I'm misunderstanding?

[accessframework]

Hi,

Yes, If you want to train the framework, you have to use the huggingface dataset, which we are planning to release soon!