biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

[compile] Recompile graph module during Dynamo cache loading. (#30743)

Browse Source 
Signed-off-by: Zhengxu Chen <zhxchen17@fb.com>
This commit is contained in:
Zhengxu Chen
2025-12-17 05:00:12 -05:00
committed by GitHub
parent 7b966ae2ba
commit 53cd7f868b
1 changed files with 1 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

1
vllm/compilation/caching.py
View File

@@ -104,6 +104,7 @@ class VllmSerializableFunction(SerializableCallable):
state = pickle.loads(data) state = pickle.loads(data)
fake_mode = FakeTensorMode(shape_env=ShapeEnv()) fake_mode = FakeTensorMode(shape_env=ShapeEnv())
state["graph_module"] = GraphPickler.loads(state["graph_module"], fake_mode) state["graph_module"] = GraphPickler.loads(state["graph_module"], fake_mode)
state["graph_module"].recompile()
state["example_inputs"] = GraphPickler.loads(state["example_inputs"], fake_mode) state["example_inputs"] = GraphPickler.loads(state["example_inputs"], fake_mode)
vllm_backend = VllmBackend(get_current_vllm_config(), state["prefix"]) vllm_backend = VllmBackend(get_current_vllm_config(), state["prefix"])