feat: support compatibility between MTP and two-batch-overlap (#7225)

Co-authored-by: Cheng Wan <54331508+ch-wan@users.noreply.github.com>
2025-06-27 01:10:27 -07:00
parent 1b95162008
commit 41650b0d70
7 changed files with 325 additions and 38 deletions
--- a/python/sglang/srt/model_executor/cuda_graph_runner.py
+++ b/python/sglang/srt/model_executor/cuda_graph_runner.py
@@ -679,6 +679,7 @@ class CudaGraphRunner:
                forward_mode=self.capture_forward_mode,
                bs=bs,
                num_token_non_padded=len(forward_batch.input_ids),
+                spec_info=forward_batch.spec_info,
            )
        if forward_batch.forward_mode.is_idle() and forward_batch.spec_info is not None:
            forward_batch.spec_info.custom_mask = self.custom_mask