disabled vllm tests that depends on anymodel

grzegorz-k-karch · kevalmorabia97 · commit b9d66a3cf381 · 2026-06-08T10:18:49.000-07:00
Signed-off-by: Grzegorz Karch &lt;gkarch@nvidia.com&gt;
diff --git a/tests/gpu_vllm/torch/puzzletron/test_calc_runtime_stats.py b/tests/gpu_vllm/torch/puzzletron/test_calc_runtime_stats.py
@@ -33,6 +33,7 @@
 pytest.importorskip("vllm", reason="vLLM is required for calc_runtime_for_subblocks")
 
 
+@pytest.mark.skip(reason="AnyModel is not supported in vLLM yet")
 def test_calc_runtime_for_subblocks(tmp_path: Path):
     """End-to-end: a tiny subblock set yields a runtime dict + positive no-block overhead."""
     tokenizer = get_tiny_tokenizer()
