benchmark

ad21edb9 - [wconstab/ltc] Make nvidia_deeprecommender accept lazy device (#643)

[wconstab/ltc] Make nvidia_deeprecommender accept lazy device (#643) This pull request makes nvidia_deeprecommender accept lazy device. It just fixes the issue in the ltc branch instead of main since I don't think it's applicable to make the model robust to unknown device.