converting potential binary file to ASCII format

[Hasan Babaei]

Hello,

I need to get the structure of the aenet potential file. I tried to compile the tool trnset2ASCII.f90 in order to convert my potential to a text file. However, I could not compile it. It seems some of the files are f77 formation and not compatible with recent release.

I was wondering if there is any way I can use to obtain the structure of the potential file. I need to convert it to another format in order to use lammps.

Best,

Hasan

[Nongnuch Artrith]

Hello Hasan, 

You can compile the trnset2ASCII.x tool with “make tools -f makefiles/Makefile...” from the main src directory.

Best,
Nong

[Hasan Babaei]

Thank you, Nong, for the quick response.

We could install trnset2ASCII.x successfully. However, when we try to convert our trained potential, an error regarding format comes out similar to below

 Converting binary format to ASCII text format.
At line 789 of file trainset.f90 (unit = 20, file = 'Si')

I just wonder if you have seen this problem before and might know a remedy.

[Nongnuch Artrith]

Hello Hasan,

I am sorry, the current version of the converter "trnset2ASCII" only works for training set files (*.scaled), but not yet for the potential files.  We do not have any tool to convert potentials yet.  You could probably extend "trnset2ASCII" for that purpose?

Best,
Nong

[Hasan Babaei]

Hello Nong,

I could convert the binary file to a text file. Thanks for your help.

I tried to understand how the weights are written. However, I could not find it by looking into the source code.

I need to know the the way vector W(:) is defined. Do you write the weights this way: (a) all weights between first node of layer 0 to all nodes in layer 1, then the other nodes of layer 0 until the bias node; or you do it this way: (b) weights between all nodes of layer 0 (including bias) to node 1 of layer 1, then the other nodes in layer 1. 

I will appreciate it if you let me know how W() is constructed. Thanks.

Hasan