Update rope.py

rope.py

@@ -57,4 +57,4 @@ def apply_rotary_emb(x, freqs_cis, position_ids, num_heads, rot_dim=None, interl
     yi = xr * sin + xi * cos
     y = torch.stack((yr, yi), dim=-1).flatten(-2)            # (B,H,T,rd)
 
-    return torch.cat([y.to(x.dtype), x_pass], dim=-1)
+    return torch.cat([y.to(x.dtype), x_pass], dim=-1)