TO DO.txt

Items I changed in NNReplayer:
1. weightSlack in repair_weights_class is not correct it is actually the bound of weight error
2. add more gurobi options by changing the Option class. enable adding any options for the solver_factory
3. add the capability to apply different weight values for samples and e_w slack variables through options and modifying cost_expr
4. change e_w cost to e_w**2
6. apply a mechanism to apply different parameter and variable bounds 
7. remove the extra binary variable defined on the output constraint
imp: 8. parameter bound should be compatible with bound of weight error (or weight slack in this version of code)
9. record opt and optimal variable logs
10. Specify output bounds in optimizer
11. options is modified
12. record opt log file
13. save the constructed model
14 mlp_get_weights() should also work with pytorch models!
15. play with gdp_formulation variable.
16. check what happens if an empty list of constraints is inputted to the repair obj
17. write a parser to extract the atchitechture and weights of an inputed model (including relu activations) << important
    a) it will change the input of NNRepair class to weight and architecture