Running Pytorch with torch.distributed.launch requires more memory?

前端 未结 0 939
青春惊慌失措
青春惊慌失措 2020-12-23 02:45

I can run my model with few GPUs perfectly (by few I mean anything less than 5). When I increased the number of GPUs, let\'s say 10, I got Cuda out of memory error:



        
相关标签:
回答
  • 消灭零回复
提交回复
热议问题