Respect user override for Gemma4 attention backend (#25547)

Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

python/sglang/srt/server_args.py

@@ -2195,12 +2195,29 @@ class ServerArgs:
             )
             self.disable_hybrid_swa_memory = True
         elif model_arch == "Gemma4ForConditionalGeneration":
-            if is_sm100_supported():
+            default_attention_backend = (
-                self.attention_backend = "trtllm_mha"
+                "trtllm_mha" if is_sm100_supported() else "triton"
+            )
+            if self.is_attention_backend_not_set():
+                self.attention_backend = default_attention_backend
+                logger.info(
+                    f"Use {self.attention_backend} as default attention backend for Gemma4"
+                )
             else:
-                self.attention_backend = "triton"
+                # If only one split backend is set, keep the other side on a
-            logger.info(
+                # Gemma4-compatible fallback instead of letting generic backend
-                f"Use {self.attention_backend} as default attention backend for Gemma4"
+                # selection choose an unsupported backend later.
+                if self.attention_backend is None:
+                    self.attention_backend = default_attention_backend
+
+            prefill_backend, decode_backend = self.get_attention_backends()
+            accepted_backends = ("trtllm_mha", "triton")
+            assert (
+                prefill_backend in accepted_backends
+                and decode_backend in accepted_backends
+            ), (
+                "Gemma4 only supports trtllm_mha or triton attention backend, "
+                f"got prefill={prefill_backend}, decode={decode_backend}"
             )
         elif model_arch == "MossVLForConditionalGeneration":
             if self.is_attention_backend_not_set():