Gradients of forces with respect to particle positions

[sigbjobo]

Hi, I am working on a project where I need to compute the gradient forces with respect to particle coordinates. Right now, I am thinking to do this numerically, but it is less elegant than doing it through DeePMD directly.

Is there a simple way to achieve this with the current version of DeePMD? And if not, is there a simple way to implement this?

[wanghan-iapcm]

Currently it is not supported by the tf.grad. One needs to to provides the hessian of the environment matrix, which is a customized op, with respect to the coordinates, and the corresponding ops involved in the back-propagation.