News for hackers. Stuff that matters.
Re:The State of Machine Learning Frameworks by rtkaratekid on Wednesday October 16, 2019 @12:57PM (#21272601)
Late reply, but it was a bug using XLA GPUs to add concurrency to the training process. Maybe someone figured it out or fixed it, but I've moved on already.
Re:The State of Machine Learning Frameworks by p1esk on Wednesday October 16, 2019 @01:41PM (#21273082)
You shouldn’t need XLA for multi GPU training. Have you tried training without it?