update

algorithms/worldmem/df_video.py

@@ -804,7 +804,7 @@ class WorldMemMinecraft(DiffusionForcingBase):
             new_c2w_mat = euler_to_camera_to_world_matrix(first_pose)
             self_memory_c2w = new_c2w_mat[None, None].to(device)
             self_frame_idx = torch.tensor([[context_frames_idx]]).to(device)
-            return first_frame.cpu(), self_frames.detach().cpu().clone(), self_poses.cpu(), self_memory_c2w.cpu(), self_frame_idx.cpu()
+            return first_frame.cpu(), self_frames.cpu().numpy(), self_poses.cpu(), self_memory_c2w.cpu(), self_frame_idx.cpu()
         else:
             last_frame = self_frames[-1].clone()
             self_poses = self_poses.to(device)