From a171e6b52dff47dc567657e7d51f641bdcb22774 Mon Sep 17 00:00:00 2001 From: shanjiaz Date: Mon, 18 May 2026 19:39:09 -0400 Subject: [PATCH] Add parallel drafting to v2 model runner unsupported features (#43010) Signed-off-by: shanjiaz --- vllm/config/vllm.py | 4 ++++ 1 file changed, 4 insertions(+) diff --git a/vllm/config/vllm.py b/vllm/config/vllm.py index a2753218f5..24eddd4a4a 100644 --- a/vllm/config/vllm.py +++ b/vllm/config/vllm.py @@ -1992,6 +1992,10 @@ class VllmConfig: elif speculative_config.method not in ("eagle", "eagle3", "mtp"): unsupported.append(f"speculative method '{speculative_config.method}'") + # V2 EagleSpeculator does not support parallel_drafting (required by PEagle) + if speculative_config.parallel_drafting: + unsupported.append("parallel drafting for speculative decoding") + if ( speculative_config.method == "eagle3" and self.parallel_config.pipeline_parallel_size > 1