Why the size of model is increased as the iteration increase

#1 
-rw-rw-r--.  1 tong.guo tong.guo   617297 9月  29 22:45 cpp_model_300
-rw-rw-r--.  1 tong.guo tong.guo   659433 9月  29 22:45 cpp_model_320
-rw-rw-r--.  1 tong.guo tong.guo   701281 9月  29 22:45 cpp_model_340
-rw-rw-r--.  1 tong.guo tong.guo   742697 9月  29 22:45 cpp_model_360
-rw-rw-r--.  1 tong.guo tong.guo   782961 9月  29 22:45 cpp_model_380
-rw-rw-r--.  1 tong.guo tong.guo   822505 9月  29 22:45 cpp_model_400
-rw-rw-r--.  1 tong.guo tong.guo   864929 9月  29 22:45 cpp_model_420
-rw-rw-r--.  1 tong.guo tong.guo   906129 9月  29 22:45 cpp_model_440
-rw-rw-r--.  1 tong.guo tong.guo   949273 9月  29 22:45 cpp_model_460
-rw-rw-r--.  1 tong.guo tong.guo   989537 9月  29 22:45 cpp_model_480
-rw-rw-r--.  1 tong.guo tong.guo  1029153 9月  29 22:45 cpp_model_500

I use CPP API to train.

I observce that the model size is bigger as the iteration number increase

#2

At every iteration a new tree is added to the model. That needs to be stored along with the trees of previous iterations so the model size increases.
Unlike other models like neural networks where the size of the model is static, gradient boosting is an additive model which means more parameters need to be represented for every iteration.