def main():
parser = OptionParser()
parser.add_option("-c",
"--conf",
dest="configure",
help="configure filename")
options, _ = parser.parse_args()
if options.configure:
conf_file = str(options.configure)
else:
print('please specify --conf configure filename')
exit(-1)
trainset_params, testset_params, net_params, solver_params = process_config(
conf_file)
trainset = TrainSet(trainset_params['data_path'],
trainset_params['sample'])
net_params['entity_num'] = trainset.entity_num
net_params['relation_num'] = trainset.relation_num
net_params['batch_size'] = trainset.record_num / int(
net_params['nbatches'])
model = TransAllModel(net_params)
model.build_graph()
pretrain = 'pre'
if not solver_params.has_key(
'pretrain_model') or solver_params['pretrain_model'] == '':
pretrain = 'nop'
if not testset_params.has_key(
'save_fld') or testset_params['save_fld'] == '':
testset_params[
'save_fld'] = 'models/TransAll_v2_{}_{}_{}_{}_{}_{}_{}_{}_{}_{}'.format(
trainset_params['data_path'].split('/')[-1],
trainset_params['sample'], net_params['embed_size_e'],
net_params['margin'], net_params['learning_rate'],
net_params['nbatches'], net_params['normed'],
net_params['activation'], net_params['opt'], pretrain)
print testset_params['save_fld']
testset_params['dataset'] = trainset_params['data_path'].split('/')[-1]
testset = TestSet(trainset_params['data_path'])
os.environ['CUDA_VISIBLE_DEVICES'] = solver_params['gpu_id']
testset_params['batch_size'] = net_params['batch_size']
testset_params['activation'] = net_params['activation']
test_model(model, testset, testset_params)
def main():
parser = OptionParser()
parser.add_option("-c", "--conf", dest="configure", help="configure filename")
options, _ = parser.parse_args()
if options.configure:
conf_file = str(options.configure)
else:
print('please specify --conf configure filename')
exit(-1)
trainset_params, testset_params, net_params, solver_params = process_config(conf_file)
trainset = TrainSet(trainset_params['data_path'], trainset_params['sample'])
net_params['entity_num'] = trainset.entity_num
net_params['relation_num'] = trainset.relation_num
net_params['batch_size'] = trainset.record_num / int(net_params['nbatches'])
model = ConvModel(net_params)
model.build_graph()
os.environ['CUDA_VISIBLE_DEVICES'] = solver_params['gpu_id']
if solver_params['phase'] == 'train':
batch_gen = trainset.batch_gen(net_params['batch_size'])
if not solver_params.has_key('pretrain_model') or solver_params['pretrain_model'] == '':
solver_params['pretrain_model'] = None
if not solver_params.has_key('save_fld'):
solver_params['save_fld'] = 'models/Conv_v1_{}_{}_{}_{}_{}_{}_{}_{}_{}'.format(
trainset_params['data_path'].split('/')[-1],
trainset_params['sample'],
net_params['embed_size'],
net_params['activation'],
net_params['channel'],
net_params['learning_rate'],
net_params['nbatches'],
net_params['normed'],
net_params['opt'])
elif solver_params['save_fld'] == '':
solver_params['save_fld'] = None
print solver_params['save_fld']
if not solver_params.has_key('summary_fld'):
solver_params['summary_fld']='graphs/Conv_v1_{}_{}_{}_{}_{}_{}_{}_{}_{}'.format(
trainset_params['data_path'].split('/')[-1],
trainset_params['sample'],
net_params['embed_size'],
net_params['activation'],
net_params['channel'],
net_params['learning_rate'],
net_params['nbatches'],
net_params['normed'],
net_params['opt'])
elif solver_params['summary_fld'] == '':
solver_params['summary_fld'] = None
train_model(model, batch_gen, solver_params)
if solver_params['save_fld']:
testset_params['save_fld'] = solver_params['save_fld']
testset_params['start'] = 1
testset_params['end'] = 1
testset_params['interval'] = solver_params['max_iter']
testset_params['dataset'] = trainset_params['data_path'].split('/')[-1]
testset = TestSet(trainset_params['data_path'], 'test')
testset_params['batch_size'] = net_params['batch_size']
if testset_params['testtype'] == 'link':
test_model_link(model, testset, testset_params)
elif testset_params['testtype'] == 'trip':
raise ValueError('Wait to finish.')
else:
raise ValueError('Undefined testtype.')
elif solver_params['phase'] == 'val':
raise ValueError('Wait to finish.')
elif solver_params['phase'] == 'test':
if not testset_params.has_key('save_fld') or testset_params['save_fld'] == '':
testset_params['save_fld'] = 'models/Conv_v1_{}_{}_{}_{}_{}_{}_{}_{}_{}'.format(
trainset_params['data_path'].split('/')[-1],
trainset_params['sample'],
net_params['embed_size'],
net_params['activation'],
net_params['channel'],
net_params['learning_rate'],
net_params['nbatches'],
net_params['normed'],
net_params['opt'])
print testset_params['save_fld']
testset_params['dataset'] = trainset_params['data_path'].split('/')[-1]
testset = TestSet(trainset_params['data_path'], 'test')
#testset = TestSet(trainset_params['data_path'], 'train')
testset_params['batch_size'] = net_params['batch_size']
if testset_params['testtype'] == 'link':
test_model_link(model, testset, testset_params)
elif testset_params['testtype'] == 'trip':
raise ValueError('Wait to finish.')
else:
raise ValueError('Undefined testtype.')
else:
raise ValueError('Undefined phase.')
def main():
parser = OptionParser()
parser.add_option("-c",
"--conf",
dest="configure",
help="configure filename")
options, _ = parser.parse_args()
if options.configure:
conf_file = str(options.configure)
else:
print('please specify --conf configure filename')
exit(-1)
trainset_params, testset_params, net_params, solver_params = process_config(
conf_file)
#trainset = TrainSet(trainset_params['data_path'], trainset_params['sample'], asym=True)
trainset = TrainSet(trainset_params['data_path'],
trainset_params['sample'])
net_params['entity_num'] = trainset.entity_num
net_params['relation_num'] = trainset.relation_num
net_params['batch_size'] = trainset.record_num / int(
net_params['nbatches'])
if solver_params['phase'] == 'train':
model = TransEModel(net_params)
model.build_graph()
os.environ['CUDA_VISIBLE_DEVICES'] = solver_params['gpu_id']
batch_gen = trainset.batch_gen(net_params['batch_size'])
if not solver_params.has_key(
'pretrain_model') or solver_params['pretrain_model'] == '':
solver_params['pretrain_model'] = None
if not solver_params.has_key('save_fld'):
solver_params[
'save_fld'] = 'models/TransE_{}_{}_{}_{}_{}_{}_{}_{}_{}'.format(
trainset_params['data_path'].split('/')[-1],
trainset_params['sample'], net_params['embed_size'],
net_params['margin'], net_params['learning_rate'],
net_params['nbatches'], net_params['normed'],
net_params['dorc'], net_params['opt'])
elif solver_params['save_fld'] == '':
solver_params['save_fld'] = None
print solver_params['save_fld']
if not solver_params.has_key('summary_fld'):
solver_params[
'summary_fld'] = 'graphs/TransE_{}_{}_{}_{}_{}_{}_{}_{}_{}'.format(
trainset_params['data_path'].split('/')[-1],
trainset_params['sample'], net_params['embed_size'],
net_params['margin'], net_params['learning_rate'],
net_params['nbatches'], net_params['normed'],
net_params['dorc'], net_params['opt'])
elif solver_params['summary_fld'] == '':
solver_params['summary_fld'] = None
solver_params['dorc'] = net_params['dorc']
train_model(model, batch_gen, solver_params)
if solver_params['save_fld']:
testset_params['save_fld'] = solver_params['save_fld']
testset_params['start'] = 1
testset_params['end'] = 1
testset_params['interval'] = solver_params['max_iter']
testset_params['dataset'] = trainset_params['data_path'].split(
'/')[-1]
testset = TestSet(trainset_params['data_path'], 'test')
testset_params['batch_size'] = net_params['batch_size']
if testset_params['testtype'] == 'link':
test_model_link(model, testset, testset_params)
elif testset_params['testtype'] == 'trip':
raise ValueError('Wait to finish.')
else:
raise ValueError('Undefined testtype.')
elif solver_params['phase'] == 'val':
raise ValueError('Wait to finish.')
elif solver_params['phase'] == 'test':
#models = TransEModel(net_params)
#models.build_graph()
#os.environ['CUDA_VISIBLE_DEVICES'] = '3'
models = []
for i in xrange(4):
with tf.device('/gpu:%d' % i):
models.append(TransEModel(net_params))
models[i].build_graph()
tf.get_variable_scope().reuse_variables()
if not testset_params.has_key(
'save_fld') or testset_params['save_fld'] == '':
testset_params[
'save_fld'] = 'models/TransE_{}_{}_{}_{}_{}_{}_{}_{}_{}'.format(
trainset_params['data_path'].split('/')[-1],
trainset_params['sample'], net_params['embed_size'],
net_params['margin'], net_params['learning_rate'],
net_params['nbatches'], net_params['normed'],
net_params['dorc'], net_params['opt'])
print testset_params['save_fld']
testset_params['dataset'] = trainset_params['data_path'].split('/')[-1]
testset = TestSet(trainset_params['data_path'], 'test')
#testset = TestSet(trainset_params['data_path'], 'train')
testset_params['batch_size'] = net_params['batch_size']
if testset_params['testtype'] == 'link':
test_model_link(models, testset, testset_params)
elif testset_params['testtype'] == 'trip':
raise ValueError('Wait to finish.')
else:
raise ValueError('Undefined testtype.')
else:
raise ValueError('Undefined phase.')