Skip to content

fix(ci): reserve VRAM for CUDA-stream logits adapter test #11209

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
furionw merged 1 commit into from
Jul 10, 2026
Merged

fix(ci): reserve VRAM for CUDA-stream logits adapter test #11209

Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
6 changes: 6 additions & 0 deletions components/src/dynamo/trtllm/tests/test_trtllm_logits_runtime.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -391,6 +391,12 @@ def test_adapter_invokes_or_logs_on_bad_shape(shape, expect_invoke, caplog):
)


# Unlike the rest of this module (CPU-only mocks, module-level
# profiled_vram_gib(0)), this test initializes a real CUDA context, so the
# GPU-parallel scheduler must reserve VRAM for it instead of packing it onto
# an already-full GPU as a zero-VRAM filler.
@pytest.mark.profiled_vram_gib(2.0)
@pytest.mark.requested_trtllm_vram_gib(2.0)
def test_adapter_enters_engine_cuda_stream():
"""When `stream_ptr` is non-null, the adapter wraps the processor
call in `torch.cuda.stream(ExternalStream(stream_ptr))` so the
Expand Down
Loading