gnn recipes bug fixes

[mnabian]

PhysicsNeMo Pull Request

Description

Checklist

• I am familiar with the Contributing Guidelines.
• New or existing tests cover these changes.
• The documentation is up to date with these changes.
• The CHANGELOG.md is up to date with these changes.
• An issue is linked to this pull request.
• If I am implementing a new model or modifying any existing model, I have followed the Models Implementation Coding Standards.

Dependencies

Review Process

All PRs are reviewed by the PhysicsNeMo team before merging.

Depending on which files are changed, GitHub may automatically assign a maintainer for review.

We are also testing AI-based code review tools (e.g., Greptile), which may add automated comments with a confidence score.
This score reflects the AI’s assessment of merge readiness and is not a qualitative judgment of your work, nor is
it an indication that the PR will be accepted / rejected.

AI-generated feedback should be reviewed critically for usefulness.
You are not required to respond to every AI comment, but they are intended to help both authors and reviewers.
Please react to Greptile comments with 👍 or 👎 to provide feedback on their accuracy.

[copy-pr-bot]

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

[mnabian]

/blossom-ci

[greptile-apps]

Greptile Summary

This PR makes bug fixes to GNN training recipes: the sintering-physics example removes TF GPU memory-growth configuration (aligned with switching to tensorflow-cpu), and the xaeronet example improves the PyG dependency installation instructions to auto-detect the user's torch/CUDA combination.

• The sintering-physics train.py replaces native PyTorch AMP (torch.amp.GradScaler / torch.autocast) with NVIDIA Apex AMP, but Apex is not listed in requirements.txt and requires a custom CUDA-aware build — this will break fp16 training for users who haven't manually built Apex.
• The xaeronet requirements.txt removes torch_scatter from pip-installable deps and moves it to a manual detection snippet in the README, and adds scikit-learn, tabulate, and matplotlib.
• Five files are missing trailing newlines.

Important Files Changed

Filename	Overview
examples/additive_manufacturing/sintering_physics/train.py	Removes native PyTorch AMP and reverts to NVIDIA Apex; Apex is not in requirements and requires a custom build, so fp16 training will silently call exit() for most users. Also removes TF GPU memory-growth config and trailing newline.
examples/additive_manufacturing/sintering_physics/inference.py	Removes TF GPU memory-growth configuration (appropriate since tensorflow-cpu is now specified) and drops the trailing newline.
examples/additive_manufacturing/sintering_physics/requirements.txt	Adds tensorflow-cpu>=2.15,<3.0 to align with the cpu-only TF usage; reasonable change. Missing trailing newline.
examples/cfd/external_aerodynamics/xaeronet/README.md	Installation instructions updated to auto-detect torch/CUDA versions for pyg-lib and torch_scatter; cleaner and more portable approach. Missing trailing newline.
examples/cfd/external_aerodynamics/xaeronet/requirements.txt	Removes torch_scatter (now installed via manual detection step) and adds scikit-learn, tabulate, matplotlib. Missing trailing newline.

_{Reviews (1): Last reviewed commit: "gnn recipes bug fixes" | Re-trigger Greptile}

[mnabian]

/blossom-ci

[ktangsali]

LGTM

[ktangsali]

Linting is passing on this. Since this is an examples only change, pytest is not required. Merging it.