Can we add code data to continue train ?

[mx8435]

Hi, Thanks for your great job. I have the following questions:
1) When used for code pretraining, what are the advantages for GLM architecture compared to GPT ？
2) Can we load GLM-130B pretrained to pretrain code corpus directly.

Thanks!