config_fname = 'config.ini'
n_runs = 10
N = 1000 # generated sample size for each trained model
subset_size = 10 # for computing DDP
sample_times = 1000 # for computing DDP
###############################################################################
# Plot diversity and quality scores
print('Plotting diversity and quality scores ...')
plt.rcParams.update({'font.size': 20})
for (dataset, function) in list_data_function:
example_name = '{}+{}'.format(dataset, function)
_, _, _, _, _, lambda0_, lambda1_, _ = read_config(
config_fname, example_name)
# Data
data_obj = getattr(datasets, dataset)(N)
data = data_obj.data
# Function
func_obj = getattr(functions, function)()
func = func_obj.evaluate
div_data = diversity_score(data, subset_size, sample_times)
qa_data = quality_score(data, func_obj)
oa_data = overall_score(data, func_obj)
list_div_lambdas = []
list_qa_lambdas = []
list_oa_lambdas = []
if __name__ == "__main__":
list_lambdas0 = [1.0, 2.0, 3.0]
list_lambdas1 = [0.1, 0.2, 0.3]
n_runs = 10
def train(lambda0, lambda1, n_runs):
for i in range(n_runs):
png_path = './trained_gan/{}_{}/{}/synthesized.png'.format(lambda0, lambda1, i)
if not os.path.exists(png_path):
os.system('python run_experiment.py train --lambda0={} --lambda1={} --id={}'.format(lambda0, lambda1, i))
else:
os.system('python run_experiment.py evaluate --lambda0={} --lambda1={} --id={}'.format(lambda0, lambda1, i))
config_fname = 'config.ini'
latent_dim, noise_dim, bezier_degree, _, _, _, _, lambda0_, lambda1_, _ = read_config(config_fname)
for lambda0 in list_lambdas0:
lambda1 = lambda1_
train(lambda0, lambda1, n_runs)
for lambda1 in list_lambdas1:
lambda0 = lambda0_
train(lambda0, lambda1, n_runs)
###############################################################################
# Plot diversity and quality scores
print('Plotting diversity and quality scores ...')
plt.rcParams.update({'font.size': 20})
# Read dataset
data_fname = './data/xs_train.npy'
X = np.load(data_fname)
import os
import argparse
from run_experiment import read_config
if __name__ == "__main__":
parser = argparse.ArgumentParser(description='Train')
parser.add_argument('mode',
type=str,
default='train',
help='train or evaluate')
args = parser.parse_args()
assert args.mode in ['train', 'evaluate']
list_models = ['GAN', 'PaDGAN']
n_runs = 10
config_fname = 'config.ini'
for model_name in list_models:
_, _, _, _, _, _, _, lambda0, lambda1, _ = read_config(config_fname)
if model_name == 'GAN':
lambda0, lambda1 = 0., 0.
for i in range(n_runs):
png_path = './trained_gan/{}_{}/{}/synthesized.png'.format(
lambda0, lambda1, i)
if not os.path.exists(png_path) or args.mode == 'evaluate':
os.system(
'python run_experiment.py {} --lambda0={} --lambda1={} --id={}'
.format(args.mode, lambda0, lambda1, i))
from matplotlib import cm
import matplotlib
import seaborn as sns
import pandas as pd
from run_experiment import read_config
from bezier_gan import BezierGAN
from simulation import evaluate
from evaluation import diversity_score
from shape_plot import plot_shape
if __name__ == "__main__":
config_fname = 'config.ini'
latent_dim, noise_dim, bezier_degree, train_steps, batch_size, disc_lr, gen_lr, lambda0_, lambda1_, save_interval = read_config(config_fname)
bounds = (0., 1.)
# Read dataset
data_fname = './data/xs_train.npy'
X = np.load(data_fname)
N = X.shape[0]
list_models = ['GAN', 'PaDGAN']
###############################################################################
# Plot diversity and quality scores
print('Plotting diversity and quality scores ...')
plt.rcParams.update({'font.size': 20})
n_runs = 10