minke.yu/sglang
1
0
Fork 0
Code Issues Pull Requests Actions 12 Packages Projects Releases Wiki Activity

Fix stale CUDA graph benchmark and docs refs (#28041)

Browse Source
This commit is contained in:
Mohammad Miadh Angkad
2026-06-13 21:51:42 -07:00
committed by GitHub
parent 5da3b37a9d
commit 91c63aeb4d
4 changed files with 11 additions and 11 deletions
Download Patch File Download Diff File Expand all files Collapse all files
python/sglang/srt/model_executor/runner_backend_utils/breakable_cuda_graph/__init__.py
+2 -2
View File
@@ -3,16 +3,16 @@
Public API (also reachable via the deeper module paths):
- BreakableCUDAGraph, BreakableCUDAGraphCapture — capture/replay
- eager_on_graph — decorator that marks a callable as a graph break
- break_graph — helper that inserts a bare graph break
- enable_breakable_cuda_graph — context that flips the Breakable runtime flag
- is_in_breakable_cuda_graph — runtime flag getter
The legacy model_executor/breakable_cuda_graph/ package is a
backwards-compat shim that re-exports from here.
"""
from sglang.srt.model_executor.runner_backend_utils.breakable_cuda_graph.breakable_cuda_graph import ( # noqa: F401
BreakableCUDAGraph,
BreakableCUDAGraphCapture,
break_graph,
eager_on_graph,
)
from sglang.srt.model_executor.runner_backend_utils.breakable_cuda_graph.context import ( # noqa: F401