Replies: 2 comments 3 replies
-
I suspect the training data generated by AIMD does not cover the NEMD and EMD calculations. please find https://docs.deepmodeling.com/projects/deepmd/en/r2/troubleshooting/precision.html#enough-data |
Beta Was this translation helpful? Give feedback.
1 reply
-
I have checked the introduction of each descriptor, but I did not find any impact of these descriptors on the effect of the deep potential. I want to know will different descriptors affect the accuracy of the model and What are the differences between these descriptors? |
Beta Was this translation helpful? Give feedback.
2 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
Dear developer,
I recently used dp potential to calculate some material properties, such as thermal conductivity, and I used lammps to calculate these properties. First, the data I trained came from the AIMD of vasp. I used the trained dp potential to calculate the thermal conductivity (using both the NEMD and EMD methods) and found that there was a difference with the actual calculation. Please tell me how I can solve this problem and whether I need to change a model for training or which model is suitable for calculating the properties of materials? Hope you can give some suggestions.
Thanks。
Below is my json file and in.lammps file,
RbReSI.json
RbReSI-heatflux.txt
Beta Was this translation helpful? Give feedback.
All reactions