Closed gregoryprice12 closed 2 weeks ago
MR1319
commented
2 weeks ago Hi @gregoryprice12 ,
evaluate function is part of the model.py file and i tried replicating it on my system and i can run it without any errors.
For example if you are running it on windows "env_name\Lib\site-packages\TorchSisso\model.py"
gregoryprice12
commented
2 weeks ago Hi @MR1319 , Thanks for getting back to me so quickly. I had a look at my model.py and I dont seem to see the class or function:
""" Created on Wed May 17 09:22:50 2023
@author: muthyala.7 """
from .import FeatureSpaceConstruction as fc from .import DimensionalFeatureSpaceConstruction as dfc from .Regressor import Regressor from .Regressor_dimension import Regressor
import sys import time import pdb import numpy as np import pandas as pd import time from sympy import symbols from sklearn.model_selection import train_test_split
class SissoModel:
def init(self,data,operators=None,multi_task = None,n_expansion=None,n_term=None,k=20,device='cpu',use_gpu = False, relational_units = None,initial_screening = None,dimensionality=None,output_dim = None, regressor_screening = None,custom_unary_functions = None ,custom_binary_functions=None):
self.operators = operators
if self.operators == None: sys.exit('!! Please provide the valid operator set!!')
self.df=data
self.no_of_operators = n_expansion
if self.no_of_operators == None:
print('!! Warning:: Number of feature expansions is not provided, considering default configuration!! \n')
self.no_of_operators = 3
self.device = device
if n_term == None:
print('Warning:: Number of terms in equation is not provided, considering default configuration..!! \n')
self.dimension = 3
else: self.dimension = n_term
#else:
self.sis_features = k
self.relational_units = relational_units
self.initial_screening = initial_screening
self.dimensionality = dimensionality
self.output_dim = output_dim
self.regressor_screening = regressor_screening
self.use_gpu = use_gpu
self.custom_unary_functions = custom_unary_functions
self.custom_binary_functions = custom_binary_functions
self.multi_task = multi_task
if multi_task!=None:
self.multi_task_target = multi_task[0]
self.multi_task_features = multi_task[1]def fit(self):
if self.dimensionality == None:
if self.operators==None: sys.exit('Please provide the operators set for the non dimensional Regression!!')
if self.multi_task!=None:
print('Performing MultiTask Symbolic regression!!..')
equations,rmses,r2s =[],[],[]
for i in range(len(self.multi_task_target)):
#Get the target variable and feature variabls
print('Performing symbolic regression of',i+1,'Target variables....')
list1 =[]
list1.extend([self.multi_task_target[i]]+self.multi_task_features[i])
df1 = self.df.iloc[:,list1]
x,y,names = fc.feature_space_construction(self.operators,df1,self.no_of_operators,self.device,self.initial_screening,self.custom_unary_functions,self.custom_binary_functions).feature_space()
from .Regressor import Regressor
rmse, equation,r2,final_eq = Regressor(x,y,names,self.dimension,self.sis_features,self.device,self.use_gpu).regressor_fit()
equations.append(final_eq)
rmses.append(rmse)
r2s.append(r2)
if i+1 == len(self.multi_task_target):
print('Equations found::',equations)
return rmses, equations, r2s
else: continue
else:
x,y,names = fc.feature_space_construction(self.operators,self.df,self.no_of_operators,self.device,self.initial_screening,self.custom_unary_functions,self.custom_binary_functions).feature_space()
from .Regressor import Regressor
rmse, equation,r2,final_eq = Regressor(x,y,names,self.dimension,self.sis_features,self.device,self.use_gpu).regressor_fit()
return rmse, equation, r2,final_eq
else:
if self.multi_task!=None:
print('************************************************ Performing MultiTask Symbolic regression!!..************************************************ \n')
equations,rmses,r2s =[],[],[]
for i in range(len(self.multi_task_target)):
#Get the target variable and feature variabls
print('************************************************ Performing symbolic regression of',i+1,'Target variables....************************************************ \n')
list1 =[]
list1.extend([self.multi_task_target[i]]+self.multi_task_features[i])
df1 = self.df.iloc[:,list1]
x,y,names,dim = dfc.feature_space_construction(self.df,self.operators,self.relational_units,self.initial_screening,self.no_of_operators,self.device,self.dimensionality).feature_expansion()
#print(names)
from .Regressor_dimension import Regressor
rmse,equation,r2 = Regressor(x,y,names,dim,self.dimension,self.sis_features,self.device,self.output_dim,self.regressor_screening,self.use_gpu).regressor_fit()
equations.append(equation)
rmses.append(rmse)
r2s.append(r2)
if i+1 == len(self.multi_task_target):
print('Equations found::',equations)
return rmses, equations, r2s
else: continue
else:
x,y,names,dim = dfc.feature_space_construction(self.df,self.operators,self.relational_units,self.initial_screening,self.no_of_operators,self.device,self.dimensionality).feature_expansion()
#print(names)
from .Regressor_dimension import Regressor
rmse,equation,r2 = Regressor(x,y,names,dim,self.dimension,self.sis_features,self.device,self.output_dim,self.regressor_screening,self.use_gpu).regressor_fit()
return rmse,equation,r2Maybe I am missing something, Thanks, Greg
gregoryprice12
commented
2 weeks ago Hi I found the issue. I had a version 0.1.6 rather than the 0.1.8 in the example scripts. After updating it works great now. Thanks for your help.
Hi, really great package. I've just been going through the examples for train/test in the colab example and they work. When I install the package on my machine and then try to repeat I seem to be missing the evaluate class/function. I can't see it in any of the .py files installed with the package as well. Thanks, Greg