cuda_ error_ out_ of_ Memory (out of memory)

Hello everyone , I meet you again , I'm your friend, Quan Jun .

【 Said in the previous 】 Complex machine learning programs , Especially when you have already trained to run , It's not about changing CUDA It can be changed . Because many parameters are already in another gpu Inside the , Simply change one test Code is not enough . If this is the case , Only check the operation , See what program takes up so much memory , If you can kill Just drop it .

An error is as follows

Ideas

Simple and clear , He has told you , The one used by default gpu Out of memory .

Enter the following in the operating system , Check it out memory The present state ：

nvidia-smi

Copy

harm , Find out GPU-0 There is a process in progress that leads to 1GB The rest is not enough . We use it GPU-1 Just execute ！

Problem solving

python In file ：

import os

Copy

os.environ["CUDA_VISIBLE_DEVICES"] = '1'

Copy

It's solved .

Publisher ： Full stack programmer stack length , Reprint please indicate the source ：https://javaforall.cn/127682.html Link to the original text ：https://javaforall.cn