Make Q contiguous before TRT-LLM MHA decode (#31667)

2026-07-20 15:21:00 -07:00
parent 7fe9ad25ac
commit a82ead53bd
2 changed files with 4 additions and 9 deletions
@@ -9,7 +9,6 @@ register_cuda_ci(
    est_time=300,
    stage="extra-a",
    runner_config="1-gpu-small",
-    disabled="Temporarily disabled due to failing accuracy",
 )


@@ -31,19 +30,12 @@ class TestLlama8BNVFP4KVCacheSM120(GSM8KMixin, DefaultServerBase):
    other_args = [
        "--quantization",
        "modelopt_fp4",
-        "--fp4-gemm-backend",
-        "auto",
        "--kv-cache-dtype",
        "nvfp4",
        "--prefill-attention-backend",
        "flashinfer",
        "--decode-attention-backend",
        "trtllm_mha",
-        "--page-size",
-        "64",
-        "--mem-fraction-static",
-        "0.87",
-        "--cuda-graph-backend-prefill=disabled",
    ]