A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 9, 10:00 UTC
[Misc] Remove missed `pad_for_cudagraph` (#33283)
Signed-off-by: Lucas Wilkinson <lwilkins@redhat.com>
Lucas Wilkinson committed
Jan 29, 2026 at 02:12 UTC
a650ad1588059d59fdd432089c9abaf254657fe2
1 file changed
-7
vllm/config/vllm.py
-7
@@ -358,13 +358,6 @@ class VllmConfig:
358
]
359
return hash_str
360
361
- def pad_for_cudagraph(self, batch_size: int) -> int:
362
- # if batch_size > self.compilation_config.max_cudagraph_capture_size,
363
- # it should raise an IndexError.
364
- # the caller should make sure the batch_size is within the range,
365
- # i.e., batch_size <= self.compilation_config.max_cudagraph_capture_size
366
- return self.compilation_config.bs_to_padded_graph_size[batch_size]
367
-
361
@property
362
def needs_dp_coordinator(self) -> bool:
363
"""