Remove custom pre attention scaling and use computed value instead.

convert-hf-to-gguf.py

@@ -2369,9 +2369,6 @@ class Gemma2Model(Model):
         self.gguf_writer.add_final_logit_softcapping(
             self.hparams["final_logit_softcapping"]
         )
-        self.gguf_writer.add_query_pre_attn_scalar(
-            self.hparams["query_pre_attn_scalar"]
-        )
 
     def modify_tensors(self, data_torch: Tensor, name: str, bid: int | None) -> Iterable[tuple[str, Tensor]]:
         del bid  # unusem