def datagen(self, epoch=0, print_out=1, test=0):
files = self.files
# order = range(13, 17) # Temporarily train on only a few files range(len(files))
# Randomize files after first training epoch
# if epoch:
# order = np.random.permutation(order)
# choose a random sample to train on
if not test:
order = random.sample(
list(self.train_ind),
int(self.sampling_density * len(self.train_ind)))
else:
order = self.test_ind
for f_ind in order:
if print_out:
print(files[f_ind], '\n')
(X, nbrs, resnums) = helper.get_data_arrays(files[f_ind])
# normalizing the location coordinates and bond lengths and scale type encoding
# Changed the xyz normalization from 255 to 350
if self.type_feature:
Xnorm = np.concatenate([
X[:, :, :, 0:3] / 320., X[:, :, :, 3:8],
X[:, :, :, 8:] / 10.
],
axis=3)
# only consider the location coordinates and bond lengths per molecule
else:
Xnorm = np.concatenate(
[X[:, :, :, 0:3] / 320., X[:, :, :, 8:] / 10.], axis=3)
num_frames = X.shape[0]
xt_all = np.array([])
yt_all = np.array([])
num_active_frames = random.sample(
range(num_frames), int(self.sampling_density * num_frames))
print('Datagen on the following frames', num_active_frames)
for i in num_active_frames:
if self.conv_net:
xt = Xnorm[i]
if self.nbr_type == 'relative':
xt = helper.append_nbrs_relative(
xt, nbrs[i], self.molecular_nbrs)
elif self.nbr_type == 'invariant':
xt = helper.append_nbrs_invariant(
xt, nbrs[i], self.molecular_nbrs)
else:
print('Invalid nbr_type')
exit()
yt = xt.copy()
xt = xt.reshape(xt.shape[0], 1, xt.shape[1], 1)
if self.full_conv_net:
yt = xt.copy()
else:
xt = Xnorm[i]
if self.nbr_type == 'relative':
xt = helper.append_nbrs_relative(
xt, nbrs[i], self.molecular_nbrs)
elif self.nbr_type == 'invariant':
xt = helper.append_nbrs_invariant(
xt, nbrs[i], self.molecular_nbrs)
else:
print('Invalid nbr_type')
exit()
yt = xt.copy()
if not len(xt_all):
xt_all = np.expand_dims(xt, axis=0)
yt_all = np.expand_dims(yt, axis=0)
else:
xt_all = np.append(xt_all,
np.expand_dims(xt, axis=0),
axis=0)
yt_all = np.append(yt_all,
np.expand_dims(yt, axis=0),
axis=0)
yield files[f_ind], xt_all, yt_all
return
def run(GP):
# set the seed
if GP['seed']:
np.random.seed(GP['seed'])
else:
np.random.seed(np.random.randint(10000))
# Set paths
if not os.path.isdir(GP['home_dir']):
print('Keras home directory not set')
sys.exit(0)
sys.path.append(GP['home_dir'])
# Setup loggin
args = candle.ArgumentStruct(**GP)
# set_seed(args.rng_seed)
# ext = extension_from_parameters(args)
candle.verify_path(args.save_path)
prefix = args.save_path # + ext
logfile = args.logfile if args.logfile else prefix + '.log'
candle.set_up_logger(logfile, logger, False) #args.verbose
logger.info('Params: {}'.format(GP))
import p2b1 as hf
reload(hf)
#import keras_model_utils as KEU
#reload(KEU)
#reload(p2ck)
#reload(p2ck.optimizers)
maps = hf.autoencoder_preprocess()
from keras.optimizers import SGD, RMSprop, Adam
from keras.datasets import mnist
from keras.callbacks import LearningRateScheduler, ModelCheckpoint
from keras import callbacks
from keras.layers.advanced_activations import ELU
from keras.preprocessing.image import ImageDataGenerator
# GP=hf.ReadConfig(opts.config_file)
batch_size = GP['batch_size']
learning_rate = GP['learning_rate']
kerasDefaults = candle.keras_default_config()
##### Read Data ########
import helper
(data_files, fields) = p2b1.get_list_of_data_files(GP)
# Read from local directoy
#(data_files, fields) = helper.get_local_files('/p/gscratchr/brainusr/datasets/cancer/pilot2/3k_run16_10us.35fs-DPPC.20-DIPC.60-CHOL.20.dir/')
#(data_files, fields) = helper.get_local_files('3k_run16', '/p/lscratchf/brainusr/datasets/cancer/pilot2/')
# Define datagenerator
datagen = hf.ImageNoiseDataGenerator(corruption_level=GP['noise_factor'])
# get data dimension ##
num_samples = 0
for f in data_files:
# Seperate different arrays from the data
(X, nbrs, resnums) = helper.get_data_arrays(f)
num_samples += X.shape[0]
(X, nbrs, resnums) = helper.get_data_arrays(data_files[0])
print('\nData chunk shape: ', X.shape)
molecular_hidden_layers = GP['molecular_num_hidden']
if not molecular_hidden_layers:
X_train = hf.get_data(X, case=GP['case'])
input_dim = X_train.shape[1]
else:
# computing input dimension for outer AE
input_dim = X.shape[1] * molecular_hidden_layers[-1]
print('\nState AE input/output dimension: ', input_dim)
# get data dimension for molecular autoencoder
molecular_nbrs = np.int(GP['molecular_nbrs'])
num_molecules = X.shape[1]
num_beads = X.shape[2]
if GP['nbr_type'] == 'relative':
# relative x, y, z positions
num_loc_features = 3
loc_feat_vect = ['rel_x', 'rel_y', 'rel_z']
elif GP['nbr_type'] == 'invariant':
# relative distance and angle
num_loc_features = 2
loc_feat_vect = ['rel_dist', 'rel_angle']
else:
print('Invalid nbr_type!!')
exit()
if not GP['type_bool']:
# only consider molecular location coordinates
num_type_features = 0
type_feat_vect = []
else:
num_type_features = 5
type_feat_vect = list(fields.keys())[3:8]
num_features = num_loc_features + num_type_features + num_beads
dim = np.prod([num_beads, num_features, molecular_nbrs + 1])
bead_kernel_size = num_features
molecular_input_dim = dim
mol_kernel_size = num_beads
feature_vector = loc_feat_vect + type_feat_vect + list(fields.keys())[8:]
print('\nMolecular AE input/output dimension: ', molecular_input_dim)
print(
'\nData Format:\n[Frames (%s), Molecules (%s), Beads (%s), %s (%s)]' %
(num_samples, num_molecules, num_beads, feature_vector, num_features))
### Define Model, Solver and Compile ##########
print('\nDefine the model and compile')
opt = candle.build_optimizer(GP['optimizer'], learning_rate, kerasDefaults)
model_type = 'mlp'
memo = '%s_%s' % (GP['base_memo'], model_type)
######## Define Molecular Model, Solver and Compile #########
molecular_nonlinearity = GP['molecular_nonlinearity']
len_molecular_hidden_layers = len(molecular_hidden_layers)
conv_bool = GP['conv_bool']
full_conv_bool = GP['full_conv_bool']
if conv_bool:
molecular_model, molecular_encoder = AE_models.conv_dense_mol_auto(
bead_k_size=bead_kernel_size,
mol_k_size=mol_kernel_size,
weights_path=None,
input_shape=(1, molecular_input_dim, 1),
nonlinearity=molecular_nonlinearity,
hidden_layers=molecular_hidden_layers,
l2_reg=GP['l2_reg'],
drop=float(GP['drop_prob']))
elif full_conv_bool:
molecular_model, molecular_encoder = AE_models.full_conv_mol_auto(
bead_k_size=bead_kernel_size,
mol_k_size=mol_kernel_size,
weights_path=None,
input_shape=(1, molecular_input_dim, 1),
nonlinearity=molecular_nonlinearity,
hidden_layers=molecular_hidden_layers,
l2_reg=GP['l2_reg'],
drop=float(GP['drop_prob']))
else:
molecular_model, molecular_encoder = AE_models.dense_auto(
weights_path=None,
input_shape=(molecular_input_dim, ),
nonlinearity=molecular_nonlinearity,
hidden_layers=molecular_hidden_layers,
l2_reg=GP['l2_reg'],
drop=float(GP['drop_prob']))
if GP['loss'] == 'mse':
loss_func = 'mse'
elif GP['loss'] == 'custom':
loss_func = helper.combined_loss
molecular_model.compile(
optimizer=opt,
loss=loss_func,
metrics=['mean_squared_error', 'mean_absolute_error'])
print('\nModel Summary: \n')
molecular_model.summary()
##### set up callbacks and cooling for the molecular_model ##########
drop = 0.5
mb_epochs = GP['epochs']
initial_lrate = GP['learning_rate']
epochs_drop = 1 + int(np.floor(mb_epochs / 3))
def step_decay(epoch):
global initial_lrate, epochs_drop, drop
lrate = initial_lrate * np.power(drop,
np.floor((1 + epoch) / epochs_drop))
return lrate
lr_scheduler = LearningRateScheduler(step_decay)
history = callbacks.History()
# callbacks=[history,lr_scheduler]
history_logger = candle.LoggingCallback(logger.debug)
candleRemoteMonitor = candle.CandleRemoteMonitor(params=GP)
timeoutMonitor = candle.TerminateOnTimeOut(TIMEOUT)
callbacks = [history, history_logger, candleRemoteMonitor, timeoutMonitor]
loss = 0.
#### Save the Model to disk
if GP['save_path'] != None:
save_path = GP['save_path']
if not os.path.exists(save_path):
os.makedirs(save_path)
else:
save_path = '.'
model_json = molecular_model.to_json()
with open(save_path + '/model.json', "w") as json_file:
json_file.write(model_json)
encoder_json = molecular_encoder.to_json()
with open(save_path + '/encoder.json', "w") as json_file:
json_file.write(encoder_json)
print('Saved model to disk')
#### Train the Model
if GP['train_bool']:
ct = hf.Candle_Molecular_Train(
molecular_model,
molecular_encoder,
data_files,
mb_epochs,
callbacks,
batch_size=batch_size,
nbr_type=GP['nbr_type'],
save_path=GP['save_path'],
len_molecular_hidden_layers=len_molecular_hidden_layers,
molecular_nbrs=molecular_nbrs,
conv_bool=conv_bool,
full_conv_bool=full_conv_bool,
type_bool=GP['type_bool'],
sampling_density=GP['sampling_density'])
frame_loss, frame_mse = ct.train_ac()
else:
frame_mse = []
frame_loss = []
return frame_loss, frame_mse
def run(GP):
# set the seed
if GP['seed']:
np.random.seed(7)
else:
np.random.seed(np.random.randint(10000))
# Set paths
if not os.path.isdir(GP['home_dir']):
print('Keras home directory not set')
sys.exit(0)
sys.path.append(GP['home_dir'])
import p2b1_mol_AE as hf
reload(hf)
import keras_model_utils as KEU
reload(KEU)
reload(p2ck)
reload(p2ck.optimizers)
maps = hf.autoencoder_preprocess()
from keras.optimizers import SGD, RMSprop, Adam
from keras.datasets import mnist
from keras.callbacks import LearningRateScheduler, ModelCheckpoint
from keras import callbacks
from keras.layers.advanced_activations import ELU
from keras.preprocessing.image import ImageDataGenerator
# GP=hf.ReadConfig(opts.config_file)
batch_size = GP['batch_size']
learning_rate = GP['learning_rate']
kerasDefaults = p2c.keras_default_config()
##### Read Data ########
#(data_files, fields)=p2c.get_list_of_data_files(GP)
# Read from local directoy
import helper
(data_files, fields) = helper.get_local_files(
'/p/gscratchr/brainusr/datasets/cancer/pilot2/3k_run16_10us.35fs-DPPC.20-DIPC.60-CHOL.20.dir/'
)
# Define datagenerator
datagen = hf.ImageNoiseDataGenerator(corruption_level=GP['noise_factor'])
# get data dimension ##
num_samples = 0
for f in data_files:
# Seperate different arrays from the data
(X, nbrs, resnums) = helper.get_data_arrays(f)
num_samples += X.shape[0]
(X, nbrs, resnums) = helper.get_data_arrays(data_files[0])
print(X.shape)
molecular_hidden_layers = GP['molecular_num_hidden']
if not molecular_hidden_layers:
X_train = hf.get_data(X, case=GP['case'])
input_dim = X_train.shape[1]
else:
# computing input dimension for outer AE
input_dim = X.shape[1] * molecular_hidden_layers[-1]
print('The input dimension to the State AE is ', input_dim)
# get data dimension for molecular autoencoder
molecular_nbrs = np.int(GP['molecular_nbrs'])
if not GP['type_bool']:
# only consider molecular location coordinates
dim = np.prod([X.shape[2], X.shape[3] - 5, molecular_nbrs + 1])
molecular_input_dim = dim
molecular_output_dim = dim
bead_kernel_size = X.shape[3] - 5
mol_kernel_size = 12 # (X.shape[3]-5)*X.shape[2]
else:
dim = np.prod(X.shape[2:] + (molecular_nbrs + 1, ))
molecular_input_dim = dim
bead_kernel_size = X.shape[3]
mol_kernel_size = 12 # X.shape[3]*X.shape[2]
print('The input/output dimension to the Moelecular AE is ',
molecular_input_dim)
print(
'Data Format:\n [Frames (%s), Molecules (%s), Beads (%s), %s (%s)]' %
(num_samples, X.shape[1], X.shape[2], fields.keys(), X.shape[3]))
### Define Model, Solver and Compile ##########
print('Define the model and compile')
opt = p2ck.build_optimizer(GP['optimizer'], learning_rate, kerasDefaults)
model_type = 'mlp'
memo = '%s_%s' % (GP['base_memo'], model_type)
######## Define Molecular Model, Solver and Compile #########
molecular_nonlinearity = GP['molecular_nonlinearity']
len_molecular_hidden_layers = len(molecular_hidden_layers)
conv_bool = GP['conv_bool']
if conv_bool:
print('Molecular kernel size: ', mol_kernel_size)
molecular_model, molecular_encoder = hf.conv_dense_mol_auto(
bead_k_size=bead_kernel_size,
mol_k_size=mol_kernel_size,
weights_path=None,
input_shape=(1, molecular_input_dim, 1),
nonlinearity=molecular_nonlinearity,
hidden_layers=molecular_hidden_layers,
l2_reg=GP['weight_decay'],
drop=GP['drop_prob'])
else:
molecular_model = hf.dense_auto(weights_path=None,
input_shape=(molecular_input_dim, ),
nonlinearity=molecular_nonlinearity,
hidden_layers=molecular_hidden_layers,
l2_reg=GP['weight_decay'])
molecular_model.compile(
optimizer=opt,
loss=helper.combined_loss,
metrics=['mean_squared_error', 'mean_absolute_error'])
molecular_model.summary()
##### set up callbacks and cooling for the molecular_model ##########
drop = 0.5
mb_epochs = GP['molecular_epochs']
initial_lrate = GP['learning_rate']
epochs_drop = 1 + int(np.floor(mb_epochs / 3))
def step_decay(epoch):
global initial_lrate, epochs_drop, drop
lrate = initial_lrate * np.power(drop,
np.floor((1 + epoch) / epochs_drop))
return lrate
lr_scheduler = LearningRateScheduler(step_decay)
history = callbacks.History()
# callbacks=[history,lr_scheduler]
candleRemoteMonitor = CandleRemoteMonitor(params=GP)
timeoutMonitor = TerminateOnTimeOut(TIMEOUT)
callbacks = [history, candleRemoteMonitor, timeoutMonitor]
loss = 0.
#### Save the Model to disk
if GP['save_path'] != None:
if not os.path.exists(GP['save_path']):
os.makedirs(GP['save_path'])
model_json = molecular_model.to_json()
with open(GP['save_path'] + '/model.json', "w") as json_file:
json_file.write(model_json)
print('Saved model to disk')
#### Train the Model
if GP['train_bool']:
if not str2bool(GP['cool']):
effec_epochs = GP['epochs']
ct = hf.Candle_Molecular_Train(
molecular_model,
molecular_encoder,
data_files,
mb_epochs,
callbacks,
batch_size=32,
case=GP['case'],
save_path=GP['save_path'],
len_molecular_hidden_layers=len_molecular_hidden_layers,
molecular_nbrs=molecular_nbrs,
conv_bool=conv_bool,
type_bool=GP['type_bool'])
# ct=hf.Candle_Train(datagen,model,data_files,effec_epochs,case=GP['case'])
frame_loss, frame_mse = ct.train_ac()
else:
effec_epochs = GP['epochs'] // 3
ct = hf.Candle_Train(datagen,
model,
data_files,
effec_epochs,
case=GP['case'])
loss = []
for i in range(3):
lr = GP['learning_rate'] / 10**i
ct.model.optimizer.lr.set_value(lr)
if i > 0:
ct.print_data = False
print('Cooling Learning Rate by factor of 10...')
loss.extend(ct.train_ac())
return frame_loss, frame_mse
