def run_training(reduced_dataset,
output_directory,
flavor_weights,
normalization,
nodes = None,
debug = False,
events = 1000000,
other_opt_dict = {},
):
if not os.path.isdir(output_directory):
os.mkdir(output_directory)
elif glob.glob(output_directory + '/*.root*'):
raise OverwriteError('root files found in %s' % output_directory)
if nodes is None:
nodes = (20, 10)
flags = 't'
if debug: flags += 'd'
pynn.trainNN(reduced_dataset = reduced_dataset,
output_directory = output_directory,
n_iterations = 10000,
normalization = normalization,
nodes = nodes,
flavor_weights = flavor_weights,
n_training_events_target = events,
flags = flags,
**other_opt_dict)
# with_ip3d = True
# out_dir = 'weights'
settings = dict(config.items('net'))
full_ds_name = 'reduceddataset_%s_forNN.root' % input_ds
full_path = '../reduceddatasets/' + full_ds_name
class_name = 'JetFitterNN_' + input_ds
for out_dir, ip3d_state in [('ip3d_weights',True),('no_ip3d_weights',False)]:
if not os.path.isdir(out_dir):
os.mkdir(out_dir)
elif glob.glob(out_dir + '/*'):
print "files found in %s, skipping" % out_dir
continue
pynn.trainNN(input_file = full_path,
output_class = class_name,
n_iterations = int(settings['n_iterations']),
dilution_factor = int(settings['dilution_factor']),
use_sd = False,
with_ip3d = ip3d_state,
nodes_first_layer = int(settings['nodes_1']),
nodes_second_layer = int(settings['nodes_2']),
debug = True,
output_dir = out_dir)
