We comment on the consistence of the epsilon anti-symmetric tensor adopted in [R. Banerjee and S. Kulkarni, arXiv:0707.2449] when it is generalized in the general case where $sqrt{-g} eq 1$. It is pointed out that the correct non-minimal consistent gauge and gravitational anomalies should by multiplied a factor $sqrt{-g} eq 1$. We also sketch the generalization of their work to the $sqrt{-g} eq 1$ case.