About the training resources

[terry-r123]

Hi @amorehead, I would like to know how long do you need to train on 72 v100 GPUs? And is it 30 or 50 epochs?

Thanks！

[amorehead]

@terry-r123, with 72 v100 GPUs, DeepInteract models took around 2 hours to train for approximately 40 epochs each.

[terry-r123]

@amorehead, Thanks very much!

[terry-r123]

Hi @amorehead, What is the batch size with 72 v100 GPUs for around 2 hours?

[amorehead]

With this default configuration, the effective batch size is 72.

[terry-r123]

With this default configuration, the effective batch size is 72.

Thanks! I get it!