A suggested experimental workflow is to name the clade 'environment' a reference to a location in a notebook, which can be used to keep track of experimental steps
In [1]:
from environment import ex
import clades
import pandas as pd
import os
#limit the architectures that will be generated
two_layers_max = {'type': 'range', 'bounds': [1, 2]}
max_ten_units = {'type': 'range', 'bounds': [2, 10]}
#create a new sacred object, which includes the config dictionary
n1e1p1b1_dict = ex.run(config_updates=\
{'population_size':3,\
'environment':'lab3000_n1e1p1b1',\
'max_train_time':5,\
'nb_layers':two_layers_max,\
'nb_units':max_ten_units})
#create a new clade object, passing in the config dictionary
n1e1p1b1_clade = clades.GAFC1(n1e1p1b1_dict.config)
#loading the data creates train,test, and validation sets
#and also creates a folder to store the output of clade activity
n1e1p1b1_clade.load_data()
In [2]:
n1e1p1b1_clade.current_generation
Out[2]:
In [3]:
n1e1p1b1_clade.spawn()
In [4]:
n1e1p1b1_clade.genotypes
Out[4]:
In [5]:
n1e1p1b1_clade.seed_models()
In [6]:
n1e1p1b1_clade.grow_models()
^^^verbose output of n1e1p1b1_clade.grow_models()
In [7]:
n1e1p1b1_clade.phenotypes
Out[7]:
In [8]:
n1e1p1b1_clade.select_parents()
In [9]:
n1e1p1b1_clade.parent_genes
Out[9]:
In [10]:
n1e1p1b1_clade.breed()
In [11]:
n1e1p1b1_clade.current_generation
Out[11]:
In [12]:
n1e1p1b1_clade.genotypes
Out[12]:
In [ ]:
In [ ]:
In [ ]:
In [ ]:
In [ ]:
In [ ]:
In [ ]: