[Minor] make the __init__ function of model_runner.py shorter (#4132)

2025-03-06 01:51:12 -08:00
parent fcc2e37f69
commit 98c73d71cb
2 changed files with 119 additions and 102 deletions
--- a/python/sglang/srt/model_executor/cuda_graph_runner.py
+++ b/python/sglang/srt/model_executor/cuda_graph_runner.py
@@ -427,7 +427,7 @@ class CudaGraphRunner:
            self.capture_hidden_mode = hidden_mode_from_spec_info
            self.capture()

-    def replay(self, forward_batch: ForwardBatch):
+    def replay(self, forward_batch: ForwardBatch, skip_attn_backend_init: bool = False):
        self.recapture_if_needed(forward_batch)

        raw_bs = forward_batch.batch_size