-
Hi all, here I am again. Recently I encountered the problem that in each iteration, there are only 0-4 candidates in model_devi task that were chosen to take the fp task, as shown in the figure below: The structure I used to generate the dataset and to be explored in model_devi process is shown below: The parameters were set with reference to the official CH4 case and some papers, using a descriptor net of [240,240,240], a fitting net of [25,50,100]. "sel" was determined with r_cut. And the detailed param.json, parameters for init_bulk (init_param.json), the model_devi.out, lcurve.out, and INCAR files for md and single-point calculations are attached below. Can anyone help me to find out where I need to improve? Thank you! |
Beta Was this translation helpful? Give feedback.
Replies: 1 comment 2 replies
-
Your stop_batch is only 10,000, which is too small. |
Beta Was this translation helpful? Give feedback.
Your stop_batch is only 10,000, which is too small.