Can i clear up gpu vram in colab

Question

I'm trying to use aitextgen to finetune 774M gpt 2 on a dataset. unfortunately, no matter what i do, training fails because there are only 80 mb of vram available. how can i clear the vram without restarting the runtime and maybe prevent the vram from being full?

Joyanta J. Mondal · Answer 1 · 2022-10-12T14:15:38.193

9

Another solution can be using these code snippets.

1.

!pip install numba

Then:

from numba import cuda
# all of your code and execution
cuda.select_device(0)
cuda.close()

Your problem is discussed in Tensorflow official github. https://github.com/tensorflow/tensorflow/issues/36465

Update: @alchemy reported this to be unrecoverable in terms of turning on. You can try below code.

device = cuda.get_current_device() 
device.reset()

edited Oct 12 '22 at 14:15

answered Mar 07 '22 at 14:28

Joyanta J. Mondal

888
1
8
20

well i hope it will be solved soon then, i'm tired of having to finetune 124M gpt 2 only – Blazeolmo 343 Mar 07 '22 at 15:38
3

this is unrecoverable.. use `device = cuda.get_current_device(); device.reset()` https://stackoverflow.com/a/73517694/4240654 – alchemy Oct 09 '22 at 21:49

score 0 · Answer 2 · answered Mar 06 '22 at 16:24

0

Run the command !nvidia-smi inside a notebook block.
Look for the process id for the GPU that is unnecessary for you to remove for cleaning up vram. Then run the command !kill process_id

It should help you.

answered Mar 06 '22 at 16:24

Joyanta J. Mondal

888
1
8
20

there are no processes. never – Blazeolmo 343 Mar 07 '22 at 14:14
Just to recheck, is the setting "GPU" turned on from "Notebook settings" option? Because it should not be happening iff the gpu is not being on and used. – Joyanta J. Mondal Mar 07 '22 at 14:20
No no. Don't get me wrong. And don't get offended. I am trying to help you. I already provided another solution below. You can check that. – Joyanta J. Mondal Mar 07 '22 at 14:29

Can i clear up gpu vram in colab

2 Answers2