[Hybrid] Enable spec decoding in mamba cache align mode (#33705)

Signed-off-by: huanghaoyan.hhy <huanghaoyan.hhy@alibaba-inc.com>
2026-02-14 05:02:28 +08:00
parent fd267bc7b7
commit c027541eaf
2 changed files with 10 additions and 8 deletions
--- a/vllm/model_executor/models/config.py
+++ b/vllm/model_executor/models/config.py
@@ -354,10 +354,6 @@ class MambaModelConfig(VerifyAndUpdateConfig):
                assert vllm_config.scheduler_config.enable_chunked_prefill, (
                    "Chunked prefill is required for mamba cache mode 'align'."
                )
-                assert not vllm_config.speculative_config, (
-                    "Mamba cache mode 'align' is currently not compatible "
-                    "with speculative decoding."
-                )
            logger.info(
                "Warning: Prefix caching in Mamba cache '%s' "
                "mode is currently enabled. "