def process(cf):
# Enable log file
sys.stdout = Logger(cf.log_file)
print(' ---> Init experiment: ' + cf.exp_name + ' <---')
# Create the data generators
train_gen, valid_gen, test_gen = Dataset_Generators().make(cf)
# Create the optimizer
print('\n > Creating optimizer...')
optimizer = Optimizer_Factory().make(cf)
# Build model
print('\n > Building model...')
model = Model_Factory().make(cf, optimizer)
# Create the callbacks
print('\n > Creating callbacks...')
cb = Callbacks_Factory().make(cf, valid_gen)
if cf.train_model:
# Train the model
model.train(train_gen, valid_gen, cb)
# Finish
print(' ---> Finish experiment: ' + cf.exp_name + ' <---')
def process(cf):
# Enable log file
sys.stdout = Logger(cf.log_file)
print(' ---> Init experiment: ' + cf.exp_name + ' <---')
# Create the data generators
train_gen, valid_gen, test_gen = Dataset_Generators().make(cf)
# Create the optimizer
print('\n > Creating optimizer...')
optimizer = Optimizer_Factory().make(cf)
# Build model
print('\n > Building model...')
model = Model_Factory().make(cf, optimizer)
# Create the callbacks
print('\n > Creating callbacks...')
cb = Callbacks_Factory().make(cf, valid_gen)
if cf.train_model:
# Train the model
model.train(train_gen, valid_gen, cb)
if cf.test_model:
# Compute testing metrics
if (train_gen != None):
print('metrics for training')
model.test(train_gen)
# Compute validation metrics
if (valid_gen != None):
print('metrics for validation')
model.test(valid_gen)
# Compute test metrics
print('metrics for testing')
model.test(test_gen)
if cf.pred_model:
# Compute validation metrics
#model.predict(valid_gen, tag='pred')
# Compute test metrics
model.predict(test_gen, tag='pred')
# Finish
print(' ---> Finish experiment: ' + cf.exp_name + ' <---')
def process(configuration):
# Load configuration
cf = configuration.load()
# Enable log file
sys.stdout = Logger(cf.log_file)
print(' ---> Init experiment: ' + cf.exp_name + ' <---')
# Create the data generators
train_gen, valid_gen, test_gen = Dataset_Generators().make(cf)
# Create the optimizer
print('\n > Creating optimizer...')
optimizer = Optimizer_Factory().make(cf)
# Build model
print('\n > Building model...')
model = Model_Factory().make(cf, optimizer)
# Create the callbacks
print('\n > Creating callbacks...')
cb = Callbacks_Factory().make(cf, valid_gen)
try:
if cf.train_model:
# Train the model
model.train(train_gen, valid_gen, cb)
if cf.test_model:
# Compute validation metrics
model.test(valid_gen)
# Compute test metrics
model.test(test_gen)
if cf.pred_model:
# Compute validation metrics
model.predict(valid_gen, tag='pred')
# Compute test metrics
model.predict(test_gen, tag='pred')
except KeyboardInterrupt:
# In case of early stopping, transfer the local files
do_copy = raw_input(
'\033[93m KeyboardInterrupt \nDo you want to transfer files to {} ? ([y]/n) \033[0m'
.format(cf.final_savepath))
if do_copy in ['', 'y']:
# Copy result to shared directory
configuration.copy_to_shared()
raise
# Finish
print(' ---> Finish experiment: ' + cf.exp_name + ' <---')
def process(configuration):
# Load configuration
cf = configuration.load()
with tarfile.open(os.path.join(cf.savepath, "code.tar.gz"),
mode='w:gz') as archive:
archive.add(os.getcwd(), recursive=True)
# Enable log file
sys.stdout = Logger(cf.log_file)
print(' ---> Init experiment: ' + cf.exp_name + ' <---')
# Create the data generators
if cf.dataset.data_format != 'npz':
train_gen, valid_gen, test_gen = Dataset_Generators().make(cf)
else:
train_gen, valid_gen, test_gen = None, None, None
# Create the optimizer
print('\n > Creating optimizer...')
optimizer = Optimizer_Factory().make(cf)
# Build model
print('\n > Building model...')
model = Model_Factory().make(cf, optimizer)
# Create the callbacks
print('\n > Creating callbacks...')
cb = Callbacks_Factory().make(cf, valid_gen)
try:
if cf.train_model:
# Train the model
if cf.dataset.data_format == 'npz':
model.train2(cf.dataset.load_data_func, cb)
elif cf.dataset.data_format == 'folders':
model.train(train_gen, valid_gen, cb)
else:
raise ValueError('Unknown data format')
if cf.test_model:
# Compute validation metrics
# model.test(valid_gen)
# Compute test metrics
model.test(test_gen)
if cf.pred_model:
# Compute validation metrics
model.predict(valid_gen, tag='pred')
# Compute test metrics
model.predict(test_gen, tag='pred')
except KeyboardInterrupt:
# In case of early stopping, transfer the local files
do_copy = raw_input(
'\033[93m KeyboardInterrupt \nDo you want to transfer files to {} ? ([y]/n) \033[0m'
.format(cf.final_savepath))
if do_copy in ['', 'y']:
# Copy result to shared directory
configuration.copy_to_shared()
raise
# Finish
print(' ---> Finish experiment: ' + cf.exp_name + ' <---')