Skip to content

Refactor: Move CUDA graph dispatch logic earlier #27382

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
ProExpertProg merged 2 commits into from
Nov 22, 2025
Merged

Refactor: Move CUDA graph dispatch logic earlier #27382

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
bf2feee
Refactor: Move CUDA graph dispatch logic earlier
yiz-liu Nov 14, 2025
09ee430
Merge branch 'main' into main
yiz-liu Nov 21, 2025
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
50 changes: 25 additions & 25 deletions vllm/v1/worker/gpu_model_runner.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -3727,6 +3727,31 @@ def _dummy_run(
dp_rank = self.parallel_config.data_parallel_rank
num_tokens_after_padding = int(num_tokens_across_dp[dp_rank])

# filter out the valid batch descriptor
_cg_mode, batch_descriptor = (
self.cudagraph_dispatcher.dispatch(
BatchDescriptor(
num_tokens=num_tokens_after_padding,
uniform_decode=uniform_decode,
has_lora=activate_lora and self.lora_config is not None,
)
)
if not is_profile
else (CUDAGraphMode.NONE, None)
)
if cudagraph_runtime_mode is not None:
# we allow forcing NONE when the dispatcher disagrees to support
# warm ups for cudagraph capture
assert (
cudagraph_runtime_mode == CUDAGraphMode.NONE
or cudagraph_runtime_mode == _cg_mode
), (
f"Cudagraph runtime mode mismatch at dummy_run. "
f"Expected {_cg_mode}, but got {cudagraph_runtime_mode}."
)
else:
cudagraph_runtime_mode = _cg_mode

attn_metadata: PerLayerAttnMetadata | None = None

# If force_attention is True, we always capture attention. Otherwise,
Expand Down Expand Up @@ -3801,31 +3826,6 @@ def _dummy_run(
num_tokens_after_padding, None, False
)

# filter out the valid batch descriptor
_cg_mode, batch_descriptor = (
self.cudagraph_dispatcher.dispatch(
BatchDescriptor(
num_tokens=num_tokens_after_padding,
uniform_decode=uniform_decode,
has_lora=activate_lora and self.lora_config is not None,
)
)
if not is_profile
else (CUDAGraphMode.NONE, None)
)
if cudagraph_runtime_mode is not None:
# we allow forcing NONE when the dispatcher disagrees to support
# warm ups for cudagraph capture
assert (
cudagraph_runtime_mode == CUDAGraphMode.NONE
or cudagraph_runtime_mode == _cg_mode
), (
f"Cudagraph runtime mode mismatch at dummy_run. "
f"Expected {_cg_mode}, but got {cudagraph_runtime_mode}."
)
else:
cudagraph_runtime_mode = _cg_mode

if ubatch_slices is not None:
# Adjust values to reflect a single ubatch.
# TODO(sage,lucas): this is cruft that should be addressed in
Expand Down