Piecewise Cuda Graph set default (#16331)

This commit is contained in:
Yuwei An
2026-03-02 23:18:07 +08:00
committed by GitHub
parent 468e3dc56b
commit c64274c746
34 changed files with 505 additions and 143 deletions
@@ -100,7 +100,6 @@ class TestDeepseekV3FP4PiecewiseCudaGraph(CustomTestCase):
"flashinfer_trtllm",
"--quantization",
"modelopt_fp4",
"--enable-piecewise-cuda-graph",
"--kv-cache-dtype",
"fp8_e4m3",
"--model-loader-extra-config",