fix: Do not autofill default_model_filename for Python-runtime PyTorch models

[mattwittwer]

What does the PR do?

Since #484, AutoCompleteBackendFields autofills default_model_filename: "model.pt" for every backend: "pytorch" model with an empty platform — including models using the Python-based runtime (runtime: "model.py"). For those models a sibling model.pt is the weights companion to the model.py class, and the runtime gives default_model_filename top priority, so it calls torch.jit.load() on the weights file and fails at load with PytorchStreamReader failed locating file constants.pkl.

Restores the pre-#484 guard — skip filename autofill when runtime is "model.py" — and extends it to the AOTI platform branch (model.pt2), which the Python runtime cannot load either. Output is unchanged for every other config. #484's stated goal (correct .pt2 selection for torch_aoti) is preserved and now pinned by tests.

Adds AutoCompleteBackendFieldsTest to repo_agent_test (which already compiles model_config_utils.cc): the two Python-runtime regression cases plus six cases pinning existing autofill and version-dir sniffing behavior. The regression cases fail without the fix and pass with it.

Checklist

• PR title reflects the change and is of format <commit_type>: <Title>
• Changes are described in the pull request.
• Related issues are referenced.
• Populated github labels field
• Added test plan and verified test passes.
• Verified that the PR passes existing CI.
• Verified copyright is correct on all changed files.
• Added succinct git squash message before merging ref.
• All template sections are filled out.
• Optional: Additional screenshots for behavior/output changes with before/after.

Commit Type:

Check the conventional commit type
box here and add the label to the github PR.

• build
• ci
• docs
• feat
• fix
• perf
• refactor
• revert
• style
• test

Related PRs:

triton-inference-server/server#8913

Where should the reviewer start?

Test plan:

Test case added here

• CI Pipeline ID:

60326198

Caveats:

Background

Related Issues: (use one of the action keywords Closes / Fixes / Resolves / Relates to)

• closes GitHub issue: #xxx

[greptile-apps]

Greptile Summary

The PR narrows PyTorch model filename autofill and adds regression coverage.

• Skips default_model_filename autofill for Python-runtime PyTorch models on both LibTorch and AOTI platform paths.
• Preserves existing platform selection, explicit filenames, and default-runtime autofill behavior.
• Adds table-driven tests covering Python runtime regressions and existing version-directory inference behavior.

Confidence Score: 5/5

The PR appears safe to merge.

No blocking failure remains.

Important Files Changed

Filename	Overview
src/model_config_utils.cc	Adds a shared guard preventing filename autofill when the configured PyTorch runtime is model.py.
src/test/repo_agent_test.cc	Adds table-driven coverage for Python-runtime exclusions and existing PyTorch autofill and inference cases.

_{Reviews (5): Last reviewed commit: "fix pre-commit" | Re-trigger Greptile}

[yinggeh]

Is it possible to merge test cases into a for loop?

[mattwittwer]

Yes, updated