types = np.array(types)
np.random.shuffle(types)
atom_setting = ''
for i, j in zip(atoms, types):
atom_setting += i + j + '\n'
m = values['slope'].values[0]
b = values['intercept'].values[0]
length = m * start_temp + b
run = join(*[save, group, start_temp_str])
# Write run
functions.create_dir(run)
# Write input files for cubic
contents = incar_contents.replace('$side$', str(length))
contents = contents.replace('$lattice$', str(length / l))
contents = contents.replace('$time_step$', str(time_step))
contents = contents.replace('$seed$', str(seed))
contents = contents.replace('$temp$', str(start_temp))
contents = contents.replace('$steps$', str(hold_steps))
contents = contents.replace('$l$', str(l))
contents = contents.replace('$atom_setting$', atom_setting)
remove = ''
l_cubed = l**3
if atom_count < l_cubed:
density = 100000 # The number of points to use in fit
data_save_dir = '../data' # The data save folder
data_save_name = 'data.csv' # The data save name
save_plots = '../figures' # The figures save folder
order = 3 # The order of the spline fit
# Load Data
df = pd.read_csv(df)
df = df.dropna() # Drop bad jobs
df = df.sort_values(by=['volume', 'start_temperature', 'end_temperature'])
# Group data
groups = df.groupby(['composition', 'start_temperature', 'end_temperature'])
# Create figures directory
functions.create_dir(save_plots)
compositions = []
temperatures = []
lengths = []
length_errors = []
pressure_errors = []
for group, values in groups:
name = str(group)
print('Grouped by: '+name)
y = values['pressure'].values
x = values['volume'].values
x **= 1/3
hold_steps = 3000 # The number of hold steps
dT = sys.argv[1] # Change in temperature
min_temp = sys.argv[2] # The minimum allowable temperature hold
fits = pd.read_csv(fits) # Load TV curves
cwd = os.getcwd()
temp = float(os.path.basename(cwd)) - dT
if temp >= min_temp:
# Random integer
seed = np.random.randint(100000, 999999)
dir_name = os.path.join('../', str(temp))
functions.create_dir(dir_name)
# Read composition from name
composition = incar_name.split('/')[-1].split('_')[0]
fit = fits.loc[fits['composition'] == composition].values[0]
m = fit[1]
b = fit[2]
# Change volume
length = m * temp + b
# Open and read template
incar = open(incar_name)
incar_contents = incar.read()
incar.close()
import functions
import os
import platform
import victory
import use_functions
while True:
functions.print_menu() # выводим список меню
choice = input('Выберете номер пункта меню: ')
if choice == '1':
functions.create_dir() # создаем папку
elif choice == '2':
functions.delete_dir() # удаляем папку
elif choice == '3':
functions.copy_dir() # копируем папку
elif choice == '4':
print('Содержимое рабочей директории:')
print(os.listdir('.')) # папки и файлы рабочей дирректории
elif choice == '5':
functions.save_dir() # сохраняем файлы и папки рабочей дирректории
elif choice == '6':
functions.list_file() # файлы рабочей дирректории
elif choice == '7':
functions.list_folders() # папки рабочей дирректории
'max subsidy': SUB_MAX,
'reward_function': reward_fun,
'length_ep': LENGTH_EPISODE,
'n_ep': NUM_EPISODES,
'lr': LEARNING_RATE,
'batch_size': BATCH_SIZE,
'gamma': GAMMA,
'save_interval': SAVE_INTERVAL}
# ######################################################################################################################
# ######################################################################################################################
# ######################################################################################################################
# create saving directory and save config
SAVE_DIR = create_dir(DEBUG, LEARNING_RATE, DELTA_RESEARCH, BETA, DELTA_RESOURCE)
writer = SummaryWriter(SAVE_DIR)
torch.save(CONFIG, (SAVE_DIR / 'config'), _use_new_zipfile_serialization=False)
with open((SAVE_DIR / 'config.txt'), 'w') as file:
for key in CONFIG.keys():
file.write(str(key) + ': ' + str(CONFIG[key]) + '\n')
# create environment
env = FSCNetworkEnvAlternative(init_sup=INIT_SUPPORT, init_res=INIT_RESOURCE, ep_len=LENGTH_EPISODE, lambda_=LAMBDA,
sub_max=SUB_MAX, delta_res=DELTA_RESOURCE, beta=BETA, delta_search=DELTA_RESEARCH,
n_state_space=N_STATE_SPACE, base_impacts=BASE_IMPACTS, agg_weeks=4, save_calc=False)
print('-------------------------------- ' + str(device) + ' --------------------------------')
if device == 'cuda':
print(torch.cuda.get_device_name(0))
# dir:
opt_str = 'adam_lr%s_sgd_lr%s_epoch%d_de%d_batch%d' % (
args.lr_w, args.lr_gamma, args.max_epoch, args.decay_epoch,
args.batch_size)
loss_str = 'train_D_orig_beta%s_rho%s_lc%s_%s' % (args.beta, args.rho, args.lc,
args.lc_layer)
args.output_dir = os.path.join('output', '%s_%s' % (loss_str, opt_str))
args.pth_dir = os.path.join(args.output_dir, 'pth')
args.img_dir = os.path.join(args.output_dir, 'img')
args.gamma_dir = os.path.join(args.output_dir, 'gamma')
args.fid_buffer_dir = os.path.join(args.output_dir, 'fid_buffer')
args.do_IS = False
args.do_FID = False
create_dir(args.pth_dir), create_dir(args.img_dir), create_dir(
args.gamma_dir), create_dir(args.fid_buffer_dir)
# gpu:
torch.manual_seed(args.random_seed)
torch.cuda.manual_seed(args.random_seed)
os.environ['CUDA_VISIBLE_DEVICES'] = args.gpu
torch.backends.cudnn.deterministic = False
torch.backends.cudnn.benchmark = True
# set tf env
if args.do_IS:
_init_inception()
if args.do_FID:
inception_path = check_or_download_inception(None)
create_inception_graph(inception_path)
for group, values in groups:
print(group)
x = 1/values['end_temperature']
y = df['self_diffusion']
f = interp1d(x, y, kind=order) # Linear interpolation
xfit = np.linspace(min(x), max(x), density)
yfit = f(xfit)
if save_plots:
# Create saving directory for figures
functions.create_dir(save_plots)
save_name = join(save_plots, str(group))
# Plot glass transition curve
fig, ax = pl.subplots()
ax.plot(
x,
y,
linestyle='none',
marker='.',
label='Data',
color='b'
)
ax.plot(
parser.add_argument('--eval_batch_size', type=int, default=100)
parser.add_argument('--num_eval_imgs', type=int, default=50000)
parser.add_argument('--random_seed', type=int, default=12345)
parser.add_argument('--load-epoch', type=int, default=100)
parser.add_argument('--dir', type=str)
args = parser.parse_args()
os.environ['CUDA_VISIBLE_DEVICES'] = args.gpu
###
# for pruning models:
exp_str = args.dir
test_result_dir = os.path.join('output', exp_str, 'Samples')
create_dir(test_result_dir)
args.fid_buffer_dir = os.path.join('output', exp_str, 'fid_buffer')
args.load_path = os.path.join('output', exp_str, 'pth',
'epoch%d.pth' % args.load_epoch)
# # for original model:
# exp_str = 'sngan_cifar10_2019_10_24_12_19_30'
# test_result_dir = os.path.join('logs', exp_str, 'Samples')
# create_dir(test_result_dir)
# args.fid_buffer_dir = os.path.join('logs', exp_str, 'fid_buffer')
# args.load_path = os.path.join('logs', exp_str, 'Model', 'checkpoint_best.pth')
###
args.do_IS = True
args.do_FID = True