Skip to content

Fix cuda graph max bs capture upper bound #21005

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
Fridge003 merged 2 commits into from
Mar 31, 2026
Merged

Fix cuda graph max bs capture upper bound #21005

Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
4a336d9
Fix cuda graph max bs capture upper bound
Mar 20, 2026
265a484
Remove cuda graph max bs regression test
Mar 20, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
3 changes: 3 additions & 0 deletions python/sglang/srt/server_args.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1299,6 +1299,9 @@ def _generate_cuda_graph_batch_sizes(self):

capture_bs = [bs for bs in capture_bs if bs <= self.cuda_graph_max_bs]

if self.cuda_graph_max_bs not in capture_bs:
capture_bs.append(self.cuda_graph_max_bs)
Comment thread
weireweire marked this conversation as resolved.

return capture_bs

def _generate_piecewise_cuda_graph_tokens(self):
Expand Down
10 changes: 10 additions & 0 deletions test/registered/core/test_server_args.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -48,6 +48,16 @@ def test_pd_decode_defaults_to_round_robin(self):
self.assertEqual(server_args.load_balance_method, "round_robin")


class TestCudaGraphBatchSizes(unittest.TestCase):
def test_generate_cuda_graph_batch_sizes_includes_max_bs(self):
server_args = ServerArgs(model_path="dummy", cuda_graph_max_bs=500)

capture_bs = server_args._generate_cuda_graph_batch_sizes()

self.assertIn(500, capture_bs)
self.assertEqual(capture_bs[-1], 500)


class TestPortArgs(unittest.TestCase):
@patch("sglang.srt.server_args.get_free_port")
@patch("sglang.srt.server_args.tempfile.NamedTemporaryFile")
Expand Down
Loading