easy5 pts · first answer countsgpu-computingdistributed-training
Every process uses GPU 0
In an 8-GPU DDP job, all processes end up on GPU 0 and it runs out of memory. What was most likely missed?
Discussion 0
Join free to post in the discussion.
In an 8-GPU DDP job, all processes end up on GPU 0 and it runs out of memory. What was most likely missed?
Join free to post in the discussion.