Mllama_fix (quic#369)

Signed-off-by: Amit Raj <quic_amitraj@quicinc.com>

Author: quic-amitraj

QEfficient/transformers/models/mllama/modeling_mllama.py

@@ -55,19 +55,8 @@ class QEffMllamaRotaryEmbedding(MllamaRotaryEmbedding):
     - Add static sin/cos computations.
     """
 
-    def __init__(
-        self,
-        dim=None,
-        max_position_embeddings=2048,
-        base=10000,
-        device=None,
-        scaling_factor=1.0,
-        rope_type="default",
-        config: Optional[MllamaConfig] = None,
-    ):
+    def __init__(self, config: MllamaConfig, device=None):
         super().__init__(config=config)
-        inv_freq, self.attention_scaling = self.rope_init_fn(self.config, device, **self.rope_kwargs)
-        self.register_buffer("inv_freq", inv_freq, persistent=False)
 
         # Build here to make `torch.jit.trace` work.
         self._set_cos_sin_cache(