🔧 Optimize ROCm TensorFlow wheel declaration using uv.sources pattern

[coderabbitai]

Problem Description

The ROCm TensorFlow notebook's pyproject.toml currently declares the tensorflow-rocm wheel using a direct URL in the dependencies list, which is not the uv-native pattern and makes the dependencies section less clean.

Current Implementation

In jupyter/rocm/tensorflow/ubi9-python-3.12/pyproject.toml:

dependencies = [
    "tensorflow-rocm @ https://repo.radeon.com/rocm/manylinux/rocm-rel-6.4/tensorflow_rocm-2.18.1-cp312-cp312-manylinux_2_28_x86_64.whl#sha256=814a5e4842b0c92d63c7a0bb4df0baf51ff2db8615535d83fe8958204b840598",
    # ... other dependencies
]

Proposed Solution

Move the URL declaration to tool.uv.sources and add architecture requirements for better lock-time validation:

Step 1: Clean up dependencies

- "tensorflow-rocm @ https://repo.radeon.com/rocm/manylinux/rocm-rel-6.4/tensorflow_rocm-2.18.1-cp312-cp312-manylinux_2_28_x86_64.whl#sha256=814a5e4842b0c92d63c7a0bb4df0baf51ff2db8615535d83fe8958204b840598",
+ "tensorflow-rocm==2.18.1",

Step 2: Add uv.sources configuration

[tool.uv.sources]
tensorflow-rocm = { url = "https://repo.radeon.com/rocm/manylinux/rocm-rel-6.4/tensorflow_rocm-2.18.1-cp312-cp312-manylinux_2_28_x86_64.whl" }

Step 3: Add architecture requirements

[tool.uv]
# keep existing settings...
required-environments = [
    "sys_platform == 'linux' and platform_machine == 'x86_64' and implementation_name == 'cpython'",
]

Benefits

1. uv-Native Pattern: Follows uv's recommended approach for custom sources
2. Cleaner Dependencies: Keeps project.dependencies focused on package names and versions
3. Architecture Validation: Guarantees Linux x86_64 wheel exists at lock time
4. Better Maintainability: Separates source configuration from dependency declarations
5. Consistency: Aligns with other uv.sources usage in the repository

Files Affected

• jupyter/rocm/tensorflow/ubi9-python-3.12/pyproject.toml
• runtimes/rocm-tensorflow/ubi9-python-3.12/pyproject.toml (similar pattern)

Acceptance Criteria

• Move tensorflow-rocm URL from dependencies to tool.uv.sources
• Update dependencies entry to use version specification only
• Add required-environments constraint for Linux x86_64
• Verify lock regeneration works correctly with new configuration
• Apply same pattern to runtimes/rocm-tensorflow if applicable
• Update any related documentation or comments

References

• uv Sources Documentation
• uv Required Environments

Context Links

• PR: RHAIENG-304: add uv pyproject.tomls for multiple notebooks and runtimes across Jupyter, RStudio, and VSCode #2145
• Review Comment: RHAIENG-304: add uv pyproject.tomls for multiple notebooks and runtimes across Jupyter, RStudio, and VSCode #2145

[jiridanek]

70-72: Environments gate looks right; consider making x86_64 a requirement.

Keeping “linux + cpython” is good. Adding required-environments for x86_64 ensures resolver fails early if the ROCm wheel isn’t available.

Add alongside existing tool.uv:

+[tool.uv]
+# ...
+required-environments = [
+  "sys_platform == 'linux' and platform_machine == 'x86_64' and implementation_name == 'cpython'",
+]
Refs: uv environments vs required-environments. ([docs.astral.sh](https://docs.astral.sh/uv/concepts/resolution/?utm_source=chatgpt.com))