-
Notifications
You must be signed in to change notification settings - Fork 106
Open
Description
Problem Description
ROCm TensorFlow kustomization files are incorrectly referencing PyTorch image tags instead of TensorFlow image tags, causing deployed pods to use the wrong runtime environment. This is a systematic copy-paste error affecting multiple Python versions.
Affected Files
Python 3.11 (main branch):
jupyter/rocm/tensorflow/ubi9-python-3.11/kustomize/base/kustomization.yaml
- Current:
newTag: jupyter-pytorch-ubi9-python-3.11-2023a-20231002-4e9bbb3
- Expected:
newTag: rocm-jupyter-tensorflow-ubi9-python-3.11
Python 3.12 (PR #1259):
jupyter/rocm/tensorflow/ubi9-python-3.12/kustomize/base/kustomization.yaml
- Current:
newTag: jupyter-pytorch-ubi9-python-3.12-2023a-20231002-4e9bbb3
- Expected:
newTag: rocm-jupyter-tensorflow-ubi9-python-3.12
Impact
- StatefulSets deploy with PyTorch images instead of TensorFlow images
- TensorFlow-specific tests fail due to incorrect runtime environment
- Users get PyTorch environment when expecting TensorFlow capabilities
Root Cause
Copy-paste error from ROCm PyTorch kustomization files when creating ROCm TensorFlow variants.
Acceptance Criteria
- Fix image tag in
jupyter/rocm/tensorflow/ubi9-python-3.11/kustomize/base/kustomization.yaml
- Fix image tag in
jupyter/rocm/tensorflow/ubi9-python-3.12/kustomize/base/kustomization.yaml
- Verify correct image tags follow the pattern:
rocm-jupyter-tensorflow-ubi9-python-X.Y
- Ensure existing CI/CD pipeline tests pass with corrected tags
- Document the correct naming pattern to prevent future copy-paste errors
Implementation Notes
Reference the working ROCm PyTorch pattern:
jupyter/rocm/pytorch/ubi9-python-3.11/kustomize/base/kustomization.yaml
usesrocm-jupyter-pytorch-ubi9-python-3.11
jupyter/rocm/pytorch/ubi9-python-3.12/kustomize/base/kustomization.yaml
usesrocm-jupyter-pytorch-ubi9-python-3.12
Backlinks
Metadata
Metadata
Assignees
Labels
No labels
Type
Projects
Status
📋 Backlog