diff --git a/mlc_llm/core.py b/mlc_llm/core.py
index 642cdb8773..14fdad79a2 100644
--- a/mlc_llm/core.py
+++ b/mlc_llm/core.py
@@ -358,7 +358,7 @@ def mod_transform_before_build(
     )  # pylint: disable=not-callable
 
     if "num_attention_heads" in config and "hidden_size" in config:
-        if args.max_seq_len:
+        if args.max_seq_len != -1:
             mod = fuse_split_rotary_embedding(mod, config["num_attention_heads"], config["hidden_size"], args.max_seq_len)
         else:
             mod = fuse_split_rotary_embedding(mod, config["num_attention_heads"], config["hidden_size"])
diff --git a/mlc_llm/transform/fuse_split_rotary_embedding.py b/mlc_llm/transform/fuse_split_rotary_embedding.py
index 95bc667a47..e2dff432df 100644
--- a/mlc_llm/transform/fuse_split_rotary_embedding.py
+++ b/mlc_llm/transform/fuse_split_rotary_embedding.py
@@ -77,7 +77,6 @@ def split_rotary(
 
 def fuse_split_rotary_embedding(mod, num_attention_heads, hidden_size, max_sequence_length=2048):
     head_dim = hidden_size // num_attention_heads
-    print(f"fuse_split_rotary_embedding {max_sequence_length}")
 
     mod["split_rotary"] = get_split_rotary(num_attention_heads, head_dim, max_sequence_length)