Skip to content

Commit c55f2e0

Browse files
authored
[Qwen3VL] Add clear_grpah_opt_backend method to Qwen3VLForConditionalGeneration (#7110)
Add clear_grpah_opt_backend method that delegates to the underlying model to clear cuda graph optimization backend.
1 parent 054add2 commit c55f2e0

1 file changed

Lines changed: 4 additions & 0 deletions

File tree

fastdeploy/model_executor/models/qwen3_vl/qwen3_vl.py

Lines changed: 4 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -381,6 +381,10 @@ def forward(
381381

382382
return hidden_states
383383

384+
def clear_grpah_opt_backend(self):
385+
"""Clear graph optimization backend, the captured cuda graph will be cleaned"""
386+
self.model.clear_grpah_opt_backend(fd_config=self.fd_config)
387+
384388

385389
class Qwen3VLPretrainedModel(PretrainedModel):
386390
"""Utilities for tensor-parallel weight splitting."""

0 commit comments

Comments
 (0)