├── components Operator installation
├── operators
└── stack
├── containers
├── dev-base Build the DevSpaces container image.
├── devfile.yaml Defines limits, repos for the workspace.
└── src Example python programs to test GPU functionality.
oc apply -k components/operators
The DevSpace operator should install the DevWorkspace operator so there should be (2) operators installed.
oc get operators
NAME AGE
devspaces.openshift-operators 20d
devworkspace-operator.openshift-operators 20d
oc apply -k components/stack
The DevSpaces stack can take up to 10 minutes to populate its registries and become ready.
Look for the following pods to be ready in the devspaces project:
NAME READY STATUS RESTARTS AGE
che-gateway-55697854bf-twq47 4/4 Running 0 77s
devfile-registry-577789b564-5wbqh 1/1 Running 0 8m7s
devspaces-7679d5cfcb-8lzw4 1/1 Running 0 63s
devspaces-dashboard-65cf874b95-v9f9c 1/1 Running 0 2m
plugin-registry-746bb64444-j26hw 1/1 Running 0 7m54s
Setup python manually or run the devfile task from vscode.
python -m venv .venv
source .venv/bin/activate
pip install pip -Uq
pip install -r src/requirements.txt
Pytorch test.
python -m torch.utils.collect_env
Tensorflow 2.10 (versions >v2.10 report runtime errors)
python src/tf-beginner.py
Pytorch
python src/pytorch-gpu.py
Gradio/HuggingFace (This example takes at least 10 minutes to download model weights). At a minimum, an NIVDIA T4 is recommended.
python src/gradio-huggingface-code-llama.py
Delete the DevSpaces cluster instance and the devspaces namespace.
oc delete -k components/operators