How to run the code with one GPU

[xinghaoxinghao]

In fact, i only have one GPU.So if i want to run inference and dump images on a folder of images,what part of the code do I need to change.

[ajtao]

it already assumes 1 gpu. https://github.com/NVIDIA/semantic-segmentation/blob/main/scripts/dump_cityscapes.yml#L3

[xinghaoxinghao]

it already assumes 1 gpu. https://github.com/NVIDIA/semantic-segmentation/blob/main/scripts/dump_cityscapes.yml#L3

Yes, there are still errors after I execute the file.

[ajtao]

Hi @xinghaoxinghao.

Try to run the training command manually. i.e. run everything after "exec", so

python -m torch.distributed.launch --nproc_per_node=1 train.py --dataset cityscapes --cv0 ...

Another thing to try is to make sure your LOGDIR is defined with an absolute path.

[xinghaoxinghao]

Hi @xinghaoxinghao.

Try to run the training command manually. i.e. run everything after "exec", so

python -m torch.distributed.launch --nproc_per_node=1 train.py --dataset cityscapes --cv0 ...

Another thing to try is to make sure your LOGDIR is defined with an absolute path.

thank you I encountered such a problem after following your advice. ![Uploading image.png…]()

[xinghaoxinghao]

Hi @xinghaoxinghao.

Try to run the training command manually. i.e. run everything after "exec", so

python -m torch.distributed.launch --nproc_per_node=1 train.py --dataset cityscapes --cv0 ...

Another thing to try is to make sure your LOGDIR is defined with an absolute path.

thank you I encountered such a problem after following your advice

[ajtao]

Well it obviously looks like you have a problem with your apex installation. You're going to have to debug that yourself.

[raghu30213]

Hey I tried running with one GPU of 8GB but am running out of memory.