From 20bcbb98eb02d51d75fe557b60094d62df3d04fc Mon Sep 17 00:00:00 2001 From: Michal Guzek Date: Thu, 27 Aug 2026 13:43:32 -0700 Subject: [PATCH] [https://nvbugs/6566734][test] Unwaive test_disaggregated_qwen3_32b_fp8 and accept both greedy completions The qwen3_32b_fp8 disagg check required the exact substring "Asyncio in Python is a library" in the completion of a raw greedy prompt whose top-2 first tokens are near-tied: the model either answers the question or continues it. Post-merge builds 2886/2887 (DGX_H200) produced the question-continuation byte-identically and failed; the test was then waived. Accept either observed completion via the check's existing any-of list (same shape as PR #6301) and remove the waive. Signed-off-by: Michal Guzek --- .../integration/defs/disaggregated/test_disaggregated.py | 8 +++++++- tests/integration/test_lists/waives.txt | 1 - 2 files changed, 7 insertions(+), 2 deletions(-) diff --git a/tests/integration/defs/disaggregated/test_disaggregated.py b/tests/integration/defs/disaggregated/test_disaggregated.py index b76be7ed246e..ecdf8e001c71 100644 --- a/tests/integration/defs/disaggregated/test_disaggregated.py +++ b/tests/integration/defs/disaggregated/test_disaggregated.py @@ -617,9 +617,15 @@ def run_client_tests(example_dir, "Using `asyncio` in Python" ] elif "qwen3_32b_fp8" in test_desc: + # https://nvbugs/6566734: the greedy completion of the raw + # asyncio prompt is near-tied between answering the question + # and continuing it; both are valid non-garbage outputs. expected_strings = [ "The capital of Germany is Berlin", - "Asyncio in Python is a library" + [ + "Asyncio in Python is a library", + "I have read that it is used for asynchronous programming" + ] ] else: expected_strings = [ diff --git a/tests/integration/test_lists/waives.txt b/tests/integration/test_lists/waives.txt index 03282837778b..287c67474301 100644 --- a/tests/integration/test_lists/waives.txt +++ b/tests/integration/test_lists/waives.txt @@ -93,7 +93,6 @@ disaggregated/test_disaggregated.py::test_disaggregated_deepseek_v3_lite_fp8_tp1 disaggregated/test_disaggregated.py::test_disaggregated_deepseek_v3_lite_fp8_tp1_single_gpu_mtp[DeepSeek-V3-Lite-fp8] SKIP (https://nvbugs/6162322) disaggregated/test_disaggregated.py::test_disaggregated_deepseek_v3_lite_fp8_tp1_two_mtp[DeepSeek-V3-Lite-fp8] SKIP (https://nvbugs/6162322) disaggregated/test_disaggregated.py::test_disaggregated_genbs1[TinyLlama-1.1B-Chat-v1.0] SKIP (https://nvbugs/6162322) -disaggregated/test_disaggregated.py::test_disaggregated_qwen3_32b_fp8[Qwen3/Qwen3-32B-FP8] SKIP (https://nvbugs/6566734) disaggregated/test_disaggregated.py::test_disaggregated_stress_test[input8k-output1k-conc512-deepseek_r1_v2_fp4_stress] SKIP (https://nvbugs/6621358) disaggregated/test_disaggregated.py::test_disaggregated_stress_test[input8k-output1k-conc512-gpt_oss_120b_eagle_triton_stress] SKIP (https://nvbugs/6621362) disaggregated/test_disaggregated.py::test_disaggregated_stress_test[input8k-output1k-conc512-qwen3_5_4b_fp8_stress] SKIP (https://nvbugs/6621362)