About batch training

[Worldseer]

Hi, Ivan
You cut off for long sequences but don't say padding for short sequences, how do you do batch training if the length is inconsistent?
thanks

[sokrypton]

I believe batch size is always 1. So no need to pad.

[Worldseer]

thanks | | 曲鹏 | | ***@***.*** |

…

---- Replied Message ---- | From | Sergey ***@***.***> | | Date | 10/29/2022 13:09 | | To | ***@***.***> | | Cc | ***@***.******@***.***> | | Subject | Re: [gjoni/trRosetta] About batch training (Issue #15) | I believe batch size is always 1. So no need to pad. — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: ***@***.***>