News for hackers. Stuff that matters.

A comment on Nvidia announces RTX 2000 GPU series with ‘6x more performance’ and ray-tracing. Parent comment

Re:Nvidia announces RTX 2000 GPU series with ‘6x more perfor... by solomatov on Monday August 20, 2018 @03:23PM (#17803024)
You can always decrease a batch size to fit in smaller RAM.
[ Parent | on HN ]
Re:Nvidia announces RTX 2000 GPU series with ‘6x more perfor... by bitL on Monday August 20, 2018 @03:29PM (#17803070)
Unless you are already on a batch size 1 and TensorFlow greets you with a nice OOM message... Try some Wide-ResNet for state-of-art classification and play with its widening parameter.
[ Parent | on HN ]
Re:Nvidia announces RTX 2000 GPU series with ‘6x more perfor... by p1esk on Tuesday August 21, 2018 @12:55AM (#17807119)
Use gradient checkpointing.
[ Parent | on HN ]
Re:Nvidia announces RTX 2000 GPU series with ‘6x more perfor... by rspeer on Monday August 20, 2018 @10:31PM (#17806395)
...if the batch is the thing that doesn't fit in RAM, as opposed to the model.
[ Parent | on HN ]