fix EAGLE 2 non greedy case (#3407)

Co-authored-by: Ying Sheng <sqy1415@gmail.com>
2025-02-09 07:28:34 +08:00
parent f90db8bc07
commit fad315cb8e
4 changed files with 71 additions and 22 deletions
--- a/python/sglang/srt/speculative/eagle_worker.py
+++ b/python/sglang/srt/speculative/eagle_worker.py
@@ -185,6 +185,7 @@ class EAGLEWorker(TpModelWorker):
            self.topk,
            self.speculative_num_steps,
            self.server_args.speculative_num_draft_tokens,
+            batch.sampling_info.is_all_greedy,
        )

        # Free cache locations