minke.yu/sglang
1
0
Fork 0
Code Issues Pull Requests Actions 13 Packages Projects Releases Wiki Activity

Piecewise Cuda Graph set default (#16331)

Browse Source
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
Download Patch File Download Diff File Expand all files Collapse all files
test/manual/nightly/test_vlms_piecewise_cuda_graph.py
+1 -1
View File
@@ -137,7 +137,7 @@ class TestVLMPiecewiseCudaGraph(CustomTestCase):
"--trust-remote-code",
"--piecewise-cuda-graph-max-tokens",
"8192",
"--enable-piecewise-cuda-graph",
"--enforce-piecewise-cuda-graph",
"--tp=8",
"--piecewise-cuda-graph-compiler=eager",
"--disable-radix-cache",
test/manual/nightly/test_vlms_vit_cuda_graph.py
+1 -1
View File
@@ -140,7 +140,7 @@ class TestVLMViTCudaGraph(CustomTestCase):
other_args=[
"--mm-attention-backend",
"fa3",
"--enable-piecewise-cuda-graph",
"--enforce-piecewise-cuda-graph",
"--piecewise-cuda-graph-max-tokens",
"8192",
"--chunked-prefill-size",
test/manual/piecewise_cudagraph/test_disaggregation_piecewise_cuda_graph.py
+1 -1
View File
@@ -38,7 +38,7 @@ class TestDisaggregationPiecewiseCudaGraph(PDDisaggregationServerBase):
"prefill",
"--tp",
"1",
"--enable-piecewise-cuda-graph",
"--enforce-piecewise-cuda-graph",
]
prefill_args += cls.transfer_backend + cls.rdma_devices
cls.process_prefill = popen_launch_pd_server(