Follow the advice in CVXQUAD: How to use CVXQUAD's Pade Approximant instead of CVX's unreliable Successive Approximation for GP mode, log, exp, entr, rel_entr, kl_div, log_det, det_rootn, exponential cone. CVXQUAD's Quantum (Matrix) Entropy & Matrix Log related functions .
As I wrote before, fix the scaling. This may not be easy, but it is essential.
If the problem is still reported by CVX as infeasible, follow the advice in https://yalmip.github.io/debugginginfeasible/ (except for section 1, which does not apply to CVX).