A large failed_ratio can be found after 20 iteration dpgen process

[Guancred]

Hi, everyone,

I try to fit a dp model for sulfides and i have run the dpgen for 20 iteration, while the failed_ratio is still so larger as showing in the figure,

And the "model_devi_f_trust_lo": 0.15, "model_devi_f_trust_hi": 0.3, the corresponding max_devi_f is shown

And how to adjust the parameters for this situation?
Thanks!