What is the batch size used during training?

[Ziyi6]

Hi, I would like to train the model. The default batch size is 32 written in Apollo/look2hear/datas/musdb_moisesdb_datamodule.py , but the batch size written in the Apollo/configs/apollo.yaml is 1.

I think it's not possible that the batch size during training is 1?

[JusperLee]

The reason why the batch size in the config file is set to 1 is that our previous GPUs weren't very large, so we used that setting. If you have larger GPUs, you can adjust the batch size to a multiple of 4.