Nope but I guess a workaround would be to make a oneshot workaround-nvidia-gpu.service systemd unit file that runs the command and have the lxc autostart depend on it?
Might be something about PCI resets that running the command triggers π€·ββοΈ