diff --git a/unsloth/models/_utils.py b/unsloth/models/_utils.py
index 3ea38eab1d..2b8410032f 100644
--- a/unsloth/models/_utils.py
+++ b/unsloth/models/_utils.py
@@ -752,7 +752,7 @@ def patch_linear_scaling(
     fix_rope_function = """
     if getattr(self.config, "rope_scaling", None) is None:
         self.rotary_emb = {rope_function}(
-            self.head_dim,
+            dim = self.head_dim,
             max_position_embeddings=self.max_position_embeddings,
             base=self.rope_theta,
         )
@@ -761,7 +761,7 @@ def patch_linear_scaling(
         scaling_factor = self.config.rope_scaling["factor"]
         if scaling_type == "linear":
             self.rotary_emb = {scaled_rope_function}(
-                self.head_dim,
+                dim = self.head_dim,
                 max_position_embeddings=self.max_position_embeddings,
                 scaling_factor=scaling_factor,
                 base=self.rope_theta,
@@ -827,7 +827,7 @@ def patch_llama_rope_scaling(
     fix_rope_function = """
     if getattr(self.config, "rope_scaling", None) is None:
         self.rotary_emb = {rope_function}(
-            self.head_dim,
+            dim = self.head_dim,
             max_position_embeddings=self.max_position_embeddings,
             base=self.rope_theta,
         )
@@ -836,17 +836,17 @@ def patch_llama_rope_scaling(
         scaling_type2 = self.config.rope_scaling.get("rope_type", None)
         scaling_type = scaling_type1 if scaling_type1 is not None else scaling_type2
         scaling_factor = self.config.rope_scaling.get("factor")
-        
+
         if scaling_type == "linear":
             self.rotary_emb = {scaled_rope_function}(
-                self.head_dim,
+                dim = self.head_dim,
                 max_position_embeddings=self.max_position_embeddings,
                 scaling_factor=scaling_factor,
                 base=self.rope_theta,
             )
         elif scaling_type == "llama3":
             self.rotary_emb = {extended_rope_function}(
-                self.head_dim,
+                dim = self.head_dim,
                 max_position_embeddings=self.max_position_embeddings,
                 base=self.rope_theta,
             )
diff --git a/unsloth/models/gemma.py b/unsloth/models/gemma.py
index 6c9a57abf7..3dccf63ae4 100644
--- a/unsloth/models/gemma.py
+++ b/unsloth/models/gemma.py
@@ -205,9 +205,10 @@ class GemmaFixedRotaryEmbedding(torch.nn.Module):
     # Fixes https://github.com/huggingface/transformers/pull/28837
     # https://github.com/microsoft/DeepSpeed/issues/4932
     # The precision of RoPE buffers is not correct, so we cast to int64.
-    def __init__(self, dim, max_position_embeddings=2048, base=10000, device=None,
+    def __init__(self, dim = None, max_position_embeddings=2048, base=10000, device=None,
         config = None, # [TODO] Hack to pass in config - need to remove later
     ):
+        if config is not None: return # [TODO] Hack to pass in config - need to remove later
         super().__init__()
         self.dim = dim
         self.max_position_embeddings = max_position_embeddings
@@ -266,9 +267,10 @@ class GemmaFixedLinearScalingRotaryEmbedding(GemmaFixedRotaryEmbedding):
     # Fixes https://github.com/huggingface/transformers/pull/28837
     # https://github.com/microsoft/DeepSpeed/issues/4932
     # The precision of RoPE buffers is not correct, so we cast to int64.
-    def __init__(self, dim, max_position_embeddings=2048, base=10000, device=None, scaling_factor=1.0,
+    def __init__(self, dim = None, max_position_embeddings=2048, base=10000, device=None, scaling_factor=1.0,
         config = None, # [TODO] Hack to pass in config - need to remove later
     ):
+        if config is not None: return # [TODO] Hack to pass in config - need to remove later
         self.scaling_factor = scaling_factor
         super().__init__(dim, max_position_embeddings, base, device)
     pass
diff --git a/unsloth/models/llama.py b/unsloth/models/llama.py
index 929c324966..d043f03d1f 100644
--- a/unsloth/models/llama.py
+++ b/unsloth/models/llama.py
@@ -977,9 +977,10 @@ class LlamaRotaryEmbedding(torch.nn.Module):
     # Fixes https://github.com/huggingface/transformers/pull/28837
     # https://github.com/microsoft/DeepSpeed/issues/4932
     # The precision of RoPE buffers is not correct, so we cast to int64.
-    def __init__(self, dim, max_position_embeddings=2048, base=10000, device=None,
+    def __init__(self, dim = None, max_position_embeddings=2048, base=10000, device=None,
         config = None, # [TODO] Hack to pass in config - need to remove later
     ):
+        if config is not None: return # [TODO] Hack to pass in config - need to remove later
         super().__init__()
         self.dim = dim
         self.max_position_embeddings = max_position_embeddings
@@ -1032,9 +1033,10 @@ class LlamaLinearScalingRotaryEmbedding(LlamaRotaryEmbedding):
     # Fixes https://github.com/huggingface/transformers/pull/28837
     # https://github.com/microsoft/DeepSpeed/issues/4932
     # The precision of RoPE buffers is not correct, so we cast to int64.
-    def __init__(self, dim, max_position_embeddings=2048, base=10000, device=None, scaling_factor=1.0,
+    def __init__(self, dim = None, max_position_embeddings=2048, base=10000, device=None, scaling_factor=1.0,
         config = None, # [TODO] Hack to pass in config - need to remove later
     ):
+        if config is not None: return # [TODO] Hack to pass in config - need to remove later
         self.scaling_factor = scaling_factor
         super().__init__(dim, max_position_embeddings, base, device)
     pass
@@ -1059,9 +1061,10 @@ pass
 # See https://github.com/vllm-project/vllm/blob/main/vllm/model_executor/layers/rotary_embedding.py#L736
 # For Llama 3.1
 class LlamaExtendedRotaryEmbedding(torch.nn.Module):
-    def __init__(self, dim, max_position_embeddings=2048, base=10000, device=None,
+    def __init__(self, dim = None, max_position_embeddings=2048, base=10000, device=None,
         config = None, # [TODO] Hack to pass in config - need to remove later
     ):
+        if config is not None: return # [TODO] Hack to pass in config - need to remove later
         super().__init__()
         self.dim = dim
         self.max_position_embeddings = max_position_embeddings