Is there a way to set gpu_id for training TAO model?

[blanc9]

HI ail,

I ran the jupyer notebook script from cv_samples_v1.4.0 / yolo-v4 and yolo-v4-tiny on my 8-gpu machine.

when I ran two jobs at same time, I observed that two jobs are sharing same GPU.

I hope to run each job on different gpu, independently.

I tried use below command, but it shows same result.

CUDA_VISIBLE_DEVICES=1
tao yolo_v4 train -e $SPECS_DIR/yolo_v4.conf
-r $USER_EXPERIMENT_DIR/experiment_dir_unpruned_1019
-k $KEY
–gpus 1"

How can I set gpu_id for each job?

[Morganh]

--gpu_index <gpu_index> 

For example,

--gpu_index 0
--gpu_index 0 1