I modify the profiles, which reads as follows.
- display_name: "NODE1: GPU"
description: "TF 1.14 | CUDA 10.0 | Python 3.7"
kubespawner_override:
image: registry.baidubce.com/jupyter/containds-allr-datascience:cuda10.0.130-tf1.14
node_selector:
kubernetes.io/hostname: gpu-1
- display_name: "NODE2: GPU"
description: "TF 1.14 | CUDA 10.0 | Python 3.7"
kubespawner_override:
image: registry.baidubce.com/jupyter/containds-allr-datascience:cuda10.0.130-tf1.14
node_selector:
kubernetes.io/hostname: gpu-2
- display_name: "NODE1: GPU"
description: "TF 2.4 | Pytorch 1.8 | CUDA 11.0 | Python 3.8"
kubespawner_override:
image: registry.baidubce.com/jupyter/containds-allr-datascience:cuda11.1.1-tf2.4-pytorch1.8
node_selector:
kubernetes.io/hostname: gpu-1
- display_name: "NODE2: GPU"
description: "TF 2.4 | Pytorch 1.8 | CUDA 11.0 | Python 3.8"
kubespawner_override:
image: registry.baidubce.com/jupyter/containds-allr-datascience:cuda11.1.1-tf2.4-pytorch1.8
node_selector:
kubernetes.io/hostname: gpu-2
I created two images, CUDA10 and CUDA11. I want both images to be selected.But the truth is, no matter which one you choose, the service you create is the first one(CUDA10)
I think the configuration file is valid because the name displayed is correct. images is also OK, I have verified in the server pull.
I don’t know exactly what the problem is, I hope you can help me