For millimeter-wave large-scale Multiple-Input Multiple-Output (MIMO) systems with hybrid precoding, due to the hardware defects in the system, random phase shifts will be introduced in the received pilots, which will make the performance of traditional channel estimation algorithms become very poor. In some extreme conditions, even the channel information cannot be estimated. In order to solve this problem, this paper uses pilots from different radio frequency chains to have the same phase offset in the same time period, and pilots in different time periods have different phase offset characteristics to estimate the channel information with an Iterative Reweight (IR) based Partially Coherent-Compressive Phase Retrieval (PC-CPR) (IR-PC-CPR) algorithm. The algorithm first performs phase offset compensation on the received coherent pilots, and then uses an iterative reweight (IR) algorithm to estimate channel parameters to achieve high-precision channel estimation. The simulation results show that the algorithm proposed in this paper has better performance than the PC-CPR algorithm.