diff --git a/python/sglang/srt/layers/attention/aiter_backend.py b/python/sglang/srt/layers/attention/aiter_backend.py
index 761de1e22..7c9afd310 100755
--- a/python/sglang/srt/layers/attention/aiter_backend.py
+++ b/python/sglang/srt/layers/attention/aiter_backend.py
@@ -2360,10 +2360,8 @@ class AiterAttnBackend(AttentionBackend):
 
             if layer.sliding_window_size is not None and layer.sliding_window_size > -1:
                 window_size = (layer.sliding_window_size, -1)
-                # page_table = self.token_to_kv_pool.translate_loc_from_full_to_swa(
-                #    page_table
-                # )
-                page_table = self.forward_metadata.swa_page_table
+                if self.forward_metadata.swa_page_table is not None:
+                    page_table = self.forward_metadata.swa_page_table
 
             o = mha_batch_prefill_func(
                 q.contiguous().view(-1, layer.tp_q_head_num, layer.head_dim),
@@ -2496,7 +2494,8 @@ class AiterAttnBackend(AttentionBackend):
                     and layer.sliding_window_size > -1
                 ):
                     window_size = (layer.sliding_window_size - 1, 0)
-                    page_table = self.forward_metadata.swa_page_table
+                    if self.forward_metadata.swa_page_table is not None:
+                        page_table = self.forward_metadata.swa_page_table
 
                 o = torch.empty_like(q)
 
diff --git a/python/sglang/srt/models/grok.py b/python/sglang/srt/models/grok.py
index 9a096cc13..092a92630 100644
--- a/python/sglang/srt/models/grok.py
+++ b/python/sglang/srt/models/grok.py
@@ -477,7 +477,11 @@ class Grok1DecoderLayer(nn.Module):
         self.layer_id = layer_id
         self.alt_stream = alt_stream or torch.cuda.Stream()
 
-        rope_theta = config.rope_parameters["rope_theta"]
+        rope_params = getattr(config, "rope_parameters", None)
+        if rope_params and "rope_theta" in rope_params:
+            rope_theta = rope_params["rope_theta"]
+        else:
+            rope_theta = getattr(config, "rope_theta", 10000)
         self.self_attn = Grok1Attention(
             config=config,
             hidden_size=self.hidden_size,