[CI] Use an ungated model in the DGX Spark XQA test (#61416)

Signed-off-by: lesj0610 <lesj0610@gmail.com> Co-authored-by: Wei Zhao <51183510+wzhao18@users.noreply.github.com>
2026-10-14 17:04:07 +00:00
parent 3e851a7803
commit e2e3e2b834
1 changed files with 5 additions and 1 deletions
@@ -1714,7 +1714,11 @@ def test_flashinfer_xqa_nvfp4_decode_correctness(
    from vllm.v1.attention.backends import flashinfer as fi

    seq_lens, bs, d = [40, 300, 1027], 16, 256
-    config = create_vllm_config(max_model_len=max(seq_lens), block_size=bs)
+    config = create_vllm_config(
+        model_name="Qwen/Qwen3-0.6B",
+        max_model_len=max(seq_lens),
+        block_size=bs,
+    )
    # GQA 6 at head_dim 256, as Qwen3.8-27B at TP=2.
    config.model_config.model_arch_config = replace(
        config.model_config.model_arch_config,