easy5 pts · first answer countsgpu-computinggpu-memory

Reading a CUDA OOM error

Training stops with "CUDA out of memory. Tried to allocate 2.00 GiB". What does this mean?

A.The CPU ran out of RAM while loading dataB.The disk is full and cannot save checkpointsC.The model file is corruptD.The GPU did not have 2 GiB free for a new tensor

Discussion 0

Join free to post in the discussion.

Reading a CUDA OOM error

Discussion 0

Related questions