A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 3, 09:15 UTC
Add parallel drafting to v2 model runner unsupported features (#43010)
Signed-off-by: shanjiaz <zsjwpianpian@gmail.com>
shanjiaz committed
May 18, 2026 at 19:39 UTC
a171e6b52dff47dc567657e7d51f641bdcb22774
1 file changed
+4
vllm/config/vllm.py
+4
@@ -1992,6 +1992,10 @@ class VllmConfig:
1992
elif speculative_config.method not in ("eagle", "eagle3", "mtp"):
1993
unsupported.append(f"speculative method '{speculative_config.method}'")
1994
1995
+ # V2 EagleSpeculator does not support parallel_drafting (required by PEagle)
1996
+ if speculative_config.parallel_drafting:
1997
+ unsupported.append("parallel drafting for speculative decoding")
1998
+
1999
if (
2000
speculative_config.method == "eagle3"
2001
and self.parallel_config.pipeline_parallel_size > 1