serve
serve copied to clipboard
How to distribute multi models to each gpu?
I have two models: model0,model1 and two gpus: gpu0,gpu1. I want to set model0 to gpu0,model0 to gpu1,it means that the work of model0 will always on gpu0 and model1 is on gpu1. How to make it? Is it possible to implement by serve configuration or handle.py? Could you help me?Thank you very much!
Not yet unfortunately, we are planning for it https://github.com/pytorch/serve/issues/1519
torchserve does round robin assign gpu automatically.