easy5 pts · first answer countsgpu-computingdistributed-training

Launching a multi-GPU job

What does torchrun --nproc_per_node=8 train.py do on one machine?

A.Runs train.py 8 times in a rowB.Starts one process that uses 8 CPU threadsC.Starts 8 processes with rank and world size setD.Splits train.py into 8 files

Discussion 0

Join free to post in the discussion.

Launching a multi-GPU job

Discussion 0

Related questions