# How to restrict Trainer to use certain GPUs?
**URL:**
**Author:**  [@beyond](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Fu%2Fbeyond)%5C
**Post date:** [April 26, 2022, 4:05am UTC](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Fhow-to-restrict-trainer-to-use-certain-gpus%2F17185%2F1 "2022-04-26T04:05:37Z")
I have 8 gpus on my machine, I only want to train the model with for first 4 gpus. How can I achieve that?
---
**Author:**  [@BramVanroy](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Fu%2FBramVanroy)%5C
**Post date:** [April 26, 2022, 6:56am UTC](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Fhow-to-restrict-trainer-to-use-certain-gpus%2F17185%2F2 "2022-04-26T06:56:39Z")
Try again, but add the os.environ call before you import anything else. So
```auto
import os
os.environ["CUDA_VISIBLE_DEVICES"] = "0,1,2,3"
from transformers import Trainer
...
```
If you’re using Jupyter, you can also use magic commands (again, at the top, before importing anything else):
```auto
%env CUDA_VISIBLE_DEVICES=0,1,2,3
from transformers import Trainer
...
```
---
**Author:**  [@Santosh-Gupta](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Fu%2FSantosh-Gupta)%5C
**Post date:** [February 25, 2024, 8:48am UTC](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Fhow-to-restrict-trainer-to-use-certain-gpus%2F17185%2F3 "2024-02-25T08:48:48Z")
How about in cases where you would want to have additional training steps, and increase the number of devices each time? It looks like this method would need a new instance of each script