fix up default rope dtype

wanvideo/modules/model.py

@@ -118,7 +118,7 @@ def rope_apply(x, grid_sizes, freqs):
 
         # append to collection
         output.append(x_i)
-    return torch.stack(output).float()
+    return torch.stack(output).to(x.dtype)
 
 
 class WanRMSNorm(nn.Module):
@@ -1021,7 +1021,8 @@ class WanModel(ModelMixin, ConfigMixin):
         camera_embed=None,
         unianim_data=None,
         fps_embeds=None,
-        fun_ref = None
+        fun_ref = None,
+        fun_camera=None,
     ):
         r"""
         Forward pass through the diffusion model
@@ -1073,6 +1074,10 @@ class WanModel(ModelMixin, ConfigMixin):
             for u in x
             ]
 
+        if self.control_adapter is not None and fun_camera is not None:
+            fun_camera = self.control_adapter(fun_camera)
+            x = [u + v for u, v in zip(x, fun_camera)]
+
         grid_sizes = torch.stack(
             [torch.tensor(u.shape[2:], dtype=torch.long) for u in x])