Add vibevoice/pytorch-1.5B single-device inference test config

[saiarthiraguram]

Ticket

Link to Github Issue

Problem description

microsoft/VibeVoice-1.5B was brought up single-device on n150 (inference), but it is not registered in the inference test config, so it is not exercised by the test suite. The loader for it is added in a companion tt-forge-models PR (branch sai_arthi_raguram/vibe_voice), which must land and be uplifted before this entry resolves.

What's changed

Registers VibeVoice-1.5B for single-device inference testing:

vibevoice/pytorch-1.5B-single_device-inference:
  supported_archs: ["n150"]
  status: EXPECTED_PASSING

• supported_archs: ["n150"] — bringup was verified on n150 only; restrict to that arch (matches the single-arch convention used by e.g. falcon/pytorch-3_*).
• status: EXPECTED_PASSING — the registry-driven test_all_models_torch run passes with the default PCC comparison.
• No assert_pcc / required_pcc override — default PCC assertion already passes, so there's nothing to mask.
• No test-fixture change — this model runs through the registry-driven tests/runner/test_models.py::test_all_models_torch[...] path; there is no per-model test_*.py, and the loader's ModelInfo carries no bringup_status field. The YAML status is the single source of truth.

Impact: VibeVoice-1.5B is now covered by single-device inference CI on n150.

Provenance: tt-xla cb89a724b; tt-forge-models f96d6a82a0 at bringup.

Checklist

• New/Existing tests provide coverage for changes — test_all_models_torch[vibevoice/pytorch-1.5B-single_device-inference] → 1 passed in 210.41s on n150.

Logs

bringup_steps.txt
iter_2_run.log
model_overview.md