Remove graph_pool as member of VllmBackend and argument to CUDAGraphWrapper (#23385)

Signed-off-by: Luka Govedič <ProExpertProg@users.noreply.github.com>
Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com>
Co-authored-by: ProExpertProg <11367180+ProExpertProg@users.noreply.github.com>
Co-authored-by: Luka Govedič <ProExpertProg@users.noreply.github.com>
This commit is contained in:
Copilot
2025-08-25 19:34:15 -07:00
committed by GitHub
co-authored by copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com> ProExpertProg Luka Govedič
parent 6fd45e7b8a
commit 6fad29b11b
3 changed files with 7 additions and 20 deletions
+1 -4
View File
@@ -13,7 +13,7 @@ class AbstractStaticGraphWrapper(Protocol):
"""
def __init__(self, runnable: Callable, vllm_config: VllmConfig,
runtime_mode: CUDAGraphMode, graph_pool: Any, **kwargs):
runtime_mode: CUDAGraphMode, **kwargs):
"""
Initializes the StaticGraphWrapper class with graph capturing and
execution-related configurations.
@@ -25,9 +25,6 @@ class AbstractStaticGraphWrapper(Protocol):
graph runtime. See CUDAGraphMode in vllm/config.py.
Note that only the subset enum `NONE`, `PIECEWISE` and `FULL`
are used as concrete runtime mode for cudagraph dispatching.
graph_pool (Any):
Graph memory pool handle, e.g.,
`torch.cuda.graph_pool_handle()`.
Keyword Args:
kwargs: Additional keyword arguments for platform-specific
configurations.