Support using different attention backend for draft decoding. (#14843)

python/sglang/srt/model_executor/model_runner.py

@@ -2147,6 +2147,16 @@ class ModelRunner:
 
     def _get_attention_backend(self, init_new_workspace: bool = False):
         """Init attention kernel backend."""
+        draft_attn_backend = self.server_args.speculative_draft_attention_backend
+        if self.is_draft_worker and draft_attn_backend:
+            logger.warning(
+                f"Overriding draft attention backend to {draft_attn_backend}."
+            )
+            return self._get_attention_backend_from_str(
+                draft_attn_backend,
+                init_new_workspace=init_new_workspace,
+            )
+
         self.prefill_attention_backend_str, self.decode_attention_backend_str = (
             self.server_args.get_attention_backends()
         )

python/sglang/srt/server_args.py

@@ -424,6 +424,7 @@ class ServerArgs:
     speculative_accept_threshold_acc: float = 1.0
     speculative_token_map: Optional[str] = None
     speculative_attention_mode: str = "prefill"
+    speculative_draft_attention_backend: Optional[str] = None
     speculative_moe_runner_backend: Optional[str] = None
     speculative_moe_a2a_backend: Optional[str] = None
     speculative_draft_model_quantization: Optional[str] = None
@@ -3331,6 +3332,12 @@ class ServerArgs:
             help="Attention backend for speculative decoding operations (both target verify and draft extend). Can be one of 'prefill' (default) or 'decode'.",
             default=ServerArgs.speculative_attention_mode,
         )
+        parser.add_argument(
+            "--speculative-draft-attention-backend",
+            type=str,
+            help="Attention backend for speculative decoding drafting.",
+            default=ServerArgs.speculative_draft_attention_backend,
+        )
         parser.add_argument(
             "--speculative-moe-runner-backend",
             type=str,

python/sglang/srt/speculative/draft_utils.py

@@ -18,11 +18,16 @@ class DraftBackendFactory:
         self.draft_model_runner = draft_model_runner
         self.topk = topk
         self.speculative_num_steps = speculative_num_steps
+        self.draft_attn_backend = server_args.speculative_draft_attention_backend
 
     def _create_backend(
         self, backend_name: str, backend_map: dict, error_template: str
     ):
-        backend_type = getattr(self.server_args, backend_name)
+        backend_type = (
+            self.draft_attn_backend
+            if self.draft_attn_backend
+            else getattr(self.server_args, backend_name)
+        )
         if backend_type is None:
             backend_type = self.server_args.attention_backend