def plot_from_pickle(beamBendingPkl):
"""
Generates a plot of the trial results named "best_individual.png" within the trial folder.
Copies the plot into the plots folder inside the testing parent directory
:param beamBendingPkl: file name of the .pkl file containing test results
"""
if beamBendingPkl is None: # handles missing .pkl file
return None
# read bingo pickle file and retrieve best model
archipelago = load_parallel_archipelago_from_file(beamBendingPkl)
best_ind = archipelago.get_best_individual()
L = 10.
k = 5.e-5
x = np.linspace(0., L, 64).reshape([64, 1])
yan = analytic_solution(x, k, L)
y = best_ind.evaluate_equation_at(x)
# plot the training data
X_a = np.linspace(0, 10, 5)
Y_a = analytic_solution(X_a, 5e-5, 10)
plt.figure()
plt.plot(x, yan, 'r-', label='Analytical Solution')
plt.plot(X_a, Y_a, 'gx', label='Training Data Points')
plt.plot(x, y, 'b-', label='Best GPSR Model')
plt.xlabel('x')
plt.ylabel('displacement')
plt.legend()
plt.savefig('best_individual')
plt.close()
def write_results(self, beamBendingPkl):
"""
Writes the test results to the .csv in the test parent directory
:param beamBendingPkl: Filename of the .pkl file containing test results.
"""
if beamBendingPkl is None: # handles missing .pkl file
return None
# simplify equation
archipelago = load_parallel_archipelago_from_file(beamBendingPkl)
best_ind = archipelago.get_best_individual()
polynomial = Simplification(str(best_ind)).result
# write test results to the csv
with open(self.csvPath, 'a') as csvfile:
fieldNames = [
'binary', 'fitness', 'complexity', 'generations', 'f(X_0)'
]
w = csv.DictWriter(csvfile, dialect='excel', fieldnames=fieldNames)
w.writerow({
'binary': self.binary,
'fitness': best_ind.fitness,
'complexity': best_ind.get_complexity(),
'generations': archipelago.generational_age,
'f(X_0)': polynomial
})
def main():
""" main """
print() # blank line to make terminal more legible
testPath = "/uufs/chpc.utah.edu/common/home/u1008557/tests/bnd_tdata/"
csvPath = '/uufs/chpc.utah.edu/common/home/u1008557/tests/bnd_tdata/results.csv'
create_csv(csvPath)
t_list = ['t2', 't4', 't4-new', 't8', 't16', 't32', 't64']
for t in t_list:
tPath = testPath + t + '/'
os.chdir(tPath)
for i in range(1, 31):
trialPath = '{0}trial{1}/'.format(tPath, i)
os.chdir(trialPath)
beamBendingPkl = get_pkl_input()
if beamBendingPkl is None: # handles missing .pkl file
print('pkl file missing from ' + os.getcwd())
continue
#plot_from_pickle(beamBendingPkl) # comment out when troubleshooting. Plotting takes a long time
# simplify equation
try:
archipelago = load_parallel_archipelago_from_file(
beamBendingPkl)
except:
print(
'change your python path to:\nPYTHONPATH=/uufs/chpc.utah.edu/common/home/u6019587/bin/bingo_fork:/uufs/chpc.utah.edu/common/home/u6019587/src/bingo_diffeq_tf'
)
sys.exit()
best_ind = archipelago.get_best_individual()
polynomial = str(Simplification(str(best_ind)).result)
polynomial = polynomial.replace('X_0', 'x')
# polynomial = str(best_ind) # use if polynomial is broken
# Did the test converge and exit successfully?
success = 'False'
if float(best_ind.fitness) <= 1e-7:
success = 'True'
# write test results to the csv
with open(csvPath, 'a') as csvfile:
fieldNames = [
'Training Data', 'success', 'fitness', 'generations',
'solution'
]
w = csv.DictWriter(csvfile,
dialect='excel',
fieldnames=fieldNames)
w.writerow({
'Training Data': t[1:],
'success': success,
'fitness': best_ind.fitness,
'generations': archipelago.generational_age,
'solution': polynomial
})
def test_dump_then_load_equal_procs():
island = num_island(COMM_RANK)
archipelago = ParallelArchipelago(island,
sync_frequency=10,
non_blocking=True)
file_name = "testing_pa_dump_and_load_eq.pkl"
archipelago.dump_to_file(file_name)
archipelago = \
load_parallel_archipelago_from_file(file_name)
if COMM_RANK == 0:
os.remove(file_name)
origin_proc = archipelago._island.population[0].values[0]
return mpi_assert_equal(origin_proc, COMM_RANK)
def test_dump_then_load(one_island):
archipelago = ParallelArchipelago(one_island)
archipelago.evolve(1)
file_name = "testing_pa_dump_and_load.pkl"
archipelago.dump_to_file(file_name)
archipelago.evolve(1)
archipelago = \
load_parallel_archipelago_from_file(file_name)
assert 1 == archipelago.generational_age
archipelago.evolve(1)
assert 2 == archipelago.generational_age
os.remove(file_name)
def test_dump_then_load_more_procs():
island = num_island(COMM_RANK)
archipelago = ParallelArchipelago(island,
sync_frequency=10,
non_blocking=True)
file_name = "testing_pa_dump_and_load_gt.pkl"
archipelago.dump_to_file(file_name)
_remove_proc_from_pickle(file_name)
archipelago = \
load_parallel_archipelago_from_file(file_name)
if COMM_RANK == 0:
os.remove(file_name)
origin_proc = archipelago._island.population[0].values[0]
expected_origin = COMM_RANK + 1
if COMM_RANK == COMM_SIZE - 1:
expected_origin = 1
return mpi_assert_equal(origin_proc, expected_origin)
def plot_from_pickle(self, beamBendingPkl):
"""
Generates a plot of the trial results named "best_individual.png" within the trial folder.
Copies the plot into the plots folder inside the testing parent directory
:param beamBendingPkl: file name of the .pkl file containing test results
"""
if beamBendingPkl is None: # handles missing .pkl file
raise FileNotFoundError("Missing beam_bending_#####.pkl file in " +
os.getcwd())
# read bingo pickle file and retrieve best model
archipelago = load_parallel_archipelago_from_file(beamBendingPkl)
best_ind = archipelago.get_best_individual()
L = 10.
k = 5.e-2
x = np.linspace(0., L, 64).reshape([64, 1])
yan = analytic_solution(x, k, L)
y = best_ind.evaluate_equation_at(x)
# plot the training data
X_a = np.linspace(0, 10, 5)
Y_a = analytic_solution(X_a, 5e-2, 10)
plt.figure()
plt.plot(x, yan, 'r-', label='Analytical Solution')
plt.plot(X_a, Y_a, 'gx', label='Training Data Points')
plt.plot(x, y, 'b-', label='Best GPSR Model')
plt.xlabel('x')
plt.ylabel('displacement')
plt.legend()
plt.savefig('best_individual')
plt.close()
# copy plot to 'plots' directory in testing dir
cwd = os.getcwd()
plotName = self.binary + "_" + cwd[-1]
src = cwd + "/best_individual.png"
dst = self.plotsPath + plotName + ".png"
shutil.copyfile(src, dst)
def main():
""" main """
print() # blank line to make terminal more legible
testPath = "/uufs/chpc.utah.edu/common/home/u1008557/tests/bnd_study/"
csvPath = '/uufs/chpc.utah.edu/common/home/u1008557/tests/bnd_study/results.csv'
create_csv(csvPath)
# arrays containg fitness values for each test containing a certain hyperparameter
pop0, pop1, stack0, stack1, dif0, dif1, cross0, cross1, mut0, mut1 = [], [], [], [], [], [], [], [], [], []
binaries = create_binaries()
for binary in binaries:
binaryPath = testPath + binary + '/'
os.chdir(binaryPath)
beamBendingPkl = get_pkl_input()
if beamBendingPkl is None: # handles missing .pkl file
print('pkl file missing from ' + os.getcwd())
continue
plot_from_pickle(beamBendingPkl) # comment out when troubleshooting. Plotting takes a long time
# simplify equation
archipelago = load_parallel_archipelago_from_file(beamBendingPkl)
best_ind = archipelago.get_best_individual()
polynomial = str(Simplification(str(best_ind)).result)
polynomial = polynomial.replace('X_0', 'x')
# polynomial = str(best_ind) # use if polynomial is broken
# write test results to the csv
with open(csvPath, 'a') as csvfile:
fieldNames = ['binary', 'fitness', 'generations', 'solution']
w = csv.DictWriter(csvfile, dialect='excel', fieldnames=fieldNames)
w.writerow({'binary': binary, 'fitness': best_ind.fitness,
'generations': archipelago.generational_age, 'solution': polynomial})
# sort fitness into their parameter arrays
pop, stack, dif, cross, mut = binary[:]
pop, stack, dif, cross, mut = int(pop), int(stack), int(dif), int(cross), int(mut)
fit = '{:.3e}'.format(best_ind.fitness)
if pop:
pop1.append(fit)
elif not pop:
pop0.append(fit)
if stack:
stack1.append(fit)
elif not stack:
stack0.append(fit)
if dif:
dif1.append(fit)
elif not dif:
dif0.append(fit)
if cross:
cross1.append(fit)
elif not cross:
cross0.append(fit)
if mut:
mut1.append(fit)
elif not mut:
mut0.append(fit)
# write param arrays to a csv
csv_param_path = '/uufs/chpc.utah.edu/common/home/u1008557/tests/bnd_study/param_fit.csv'
if os.path.exists(csv_param_path):
os.remove(csv_param_path)
"""
pop_size = [64, 128]
stack_size = [40, 50]
differential_weight = [0.1, 0.4]
crossover_rate = [0.4, 0.6]
mutation_rate = [0.8, 1.0]
"""
# write fitness data headers to a .csv file
with open(csv_param_path, 'a') as csv_param:
fieldsnames = ['pop 64', 'pop 128', 'stack 40', 'stack 50', 'dif 0.1', 'dif 0.4',
'cross 0.4', 'cross 0.6', 'mut 0.8', 'mut 1.0']
w = csv.DictWriter(csv_param, dialect='excel', fieldnames=fieldsnames)
w.writeheader()
for i in range(len(pop0)): # write param arrays to csv
w.writerow({'pop 64': pop0[i], 'pop 128': pop1[i],
'stack 40': stack0[i], 'stack 50': stack1[i],
'dif 0.1': dif0[i], 'dif 0.4': dif1[i],
'cross 0.4': cross0[i], 'cross 0.6': cross1[i],
'mut 0.8': mut0[i], 'mut 1.0': mut1[i]})
import pandas as pd
import numpy as np
import sys
from bingo.evolutionary_optimizers.parallel_archipelago import load_parallel_archipelago_from_file
if __name__ == "__main__":
checkpoint_num = sys.argv[1]
pickle_file = "".join(['checkpoint_', str(checkpoint_num), '.pkl'])
pickle = load_parallel_archipelago_from_file(pickle_file)
for individual in pickle.hall_of_fame:
equation = individual.get_console_string()
print(individual.get_complexity(), ': ', equation)
print('Fitness: ', individual.fitness)
