Skip to content

Fix flashinfer autotune to only wrap run_once() #19004

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

Open
ch-wan wants to merge 1 commit into
base: main
Choose a base branch
from
+8 −4
Open

Fix flashinfer autotune to only wrap run_once() #19004

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
12 changes: 8 additions & 4 deletions python/sglang/srt/model_executor/model_runner.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -152,6 +152,7 @@
MultiprocessingSerializer,
cpu_has_amx_support,
dynamic_import,
empty_context,
enable_show_time_cost,
get_available_gpu_memory,
get_cpu_ids_by_node,
Expand Down Expand Up @@ -1858,12 +1859,14 @@ def _flashinfer_autotune(self):

logger.info("Running FlashInfer autotune...")

with torch.inference_mode(), autotune():
self._dummy_run(batch_size=self.req_to_token_pool.size)
self._dummy_run(
batch_size=self.req_to_token_pool.size,
run_ctx=autotune(),
)

logger.info("FlashInfer autotune completed.")

def _dummy_run(self, batch_size: int):
def _dummy_run(self, batch_size: int, run_ctx=None):
"""Run a dummy forward pass for warmup/profiling."""
if self.is_generation:
capture_forward_mode = ForwardMode.DECODE
Expand Down Expand Up @@ -2103,7 +2106,8 @@ def run_once():

torch.get_device_module(self.device).synchronize()
self.tp_group.barrier()
run_once()
with torch.inference_mode(), run_ctx or empty_context():
run_once()

def init_device_graphs(self):
"""Capture device graphs."""
Expand Down
Loading