[Perf] Remove two operations in gdn_backend extend verify path (#22444)

python/sglang/srt/layers/attention/linear/gdn_backend.py

@@ -255,6 +255,9 @@ class GDNAttnBackend(MambaAttnBackendBase):
         decode_backend = get_linear_attn_decode_backend()
         prefill_backend = get_linear_attn_prefill_backend()
         self.kernel_dispatcher = GDNKernelDispatcher(decode_backend, prefill_backend)
+        self.verify_intermediate_state_indices = torch.arange(
+            self.req_to_token_pool.size, dtype=torch.int32, device=model_runner.device
+        )
 
     def init_forward_metadata(self, forward_batch: ForwardBatch):
         super().init_forward_metadata(forward_batch)
@@ -373,14 +376,7 @@ class GDNAttnBackend(MambaAttnBackendBase):
             intermediate_conv_window_cache = (
                 mamba_cache_params.intermediate_conv_window[0]
             )
-            has_initial_states = torch.ones(
-                seq_len // forward_batch.spec_info.draft_token_num,
-                dtype=torch.bool,
-                device=forward_batch.input_ids.device,
-            )
-            intermediate_state_indices = torch.arange(
-                cache_indices.shape[0], dtype=torch.int32, device=cache_indices.device
-            )
+            intermediate_state_indices = self.verify_intermediate_state_indices
         else:
             has_initial_states = forward_batch.extend_prefix_lens > 0