GitLab

Fails when tested thought

Will be used for comp_feats

Similar concepts get the same definition

Comment out blank lines

Use "file" when it is considered standard

int overflows are leading to issues

Should be more accurate and quicker to converge

Create set_standardized_value for ease of comparision operations (Node)
Use standardized values for initial feature space creation in the SIS step
Use dgemv instead of a for loop for comparision of features whenever possible
Create a new set of functions for comparision based on a pair of node_ptr, double
   This will set up a new MPI scheme that uses reduce_all

Duplication from resolving merge conflicts

This will make it easier for users who do not want to use the bindings

Standardized the output file and added constructors to recreate model from the output files

expose c++ class/functions to python

-Saves memory
-should allow for larger sis to get past MPI send

Made Model Nodes

Models now do not connect to the main value arrays so it will be easier to serialize in the future

Did them one after another, adding both here now

all lower rungs stored on processes, this was causing issues with set_value if not done this way

All Node classes that add members get desctructors/virtual destructors

Node and OperatorNode get virtual destructors

Features now point to a position in centeral arrays.

Recurisive solutions for all rungs above 2

Feature Creation, projection scores, and l0 normalization all parallelized

Node and Unit need to be serialized

Refactored C++ SISSO code now running properly.

Further tests are needed to further debug, need to work on memory access and bringing in new features