When a specified learning executing condition is established, a command furl injection quantity ratio (CFIQ-ratio) between two fuel injectors is compulsorily changed and a fuel injection quantity error of each fuel injector is learned respectively based on the CFIQ-ratio and an air-fuel-ratio feedback correction value. Based on the learning value of fuel injection quantity error, a fuel injection period of each fuel injector is respectively corrected, whereby each fuel injection quantity error of two fuel injectors is respectively corrected with respect to each cylinder. Thereby, a ratio of fuel injection quantity between two fuel injectors is accurately controlled.