def cli(args):
logging.basicConfig(
format='%(asctime)s - %(levelname)s - %(name)s - %(message)s',
datefmt='%m/%d/%Y %H:%M:%S',
level=logging.INFO)
utils.report_gpu_info()
config = cli_config_and_parse_args(args)
print(config.__dict__)
print("Continue? Starts in 10s. [Ctrl-C] to stop.")
# let's save some seconds
select.select([sys.stdin], [], [], 3)
utils.seed_modules(
config,
numpy_seed=10000,
torch_seed=20000,
torcu_cuda_seed_all=2192,
cuda_deterministic=True,
kgegrok_base_seed=30000,
cuda_benchmark=config.cudnn_benchmark)
triple_source = data.TripleSource(config.data_dir, config.triple_order,
config.triple_delimiter)
model_class = utils.load_class_from_module(config.model, 'kgegrok.models',
'kgegrok.text_models')
evaluator = evaluation.ParallelEvaluator(config, triple_source)
# maybe roughly 10s now
select.select([sys.stdin], [], [], 4)
if config.mode == 'train':
optimizer_class = utils.load_class_from_module(config.optimizer,
'torch.optim')
cli_train(triple_source, config, model_class, optimizer_class)
elif config.mode == 'train_validate':
optimizer_class = utils.load_class_from_module(config.optimizer,
'torch.optim')
cli_train_and_validate(triple_source, config, model_class,
optimizer_class, evaluator)
elif config.mode == 'test':
cli_test(triple_source, config, model_class, evaluator)
elif config.mode == 'demo_prediction':
cli_demo_prediction(triple_source, config, model_class)
elif config.mode == 'profile':
optimizer_class = utils.load_class_from_module(config.optimizer,
'torch.optim')
cli_profile(triple_source, config, model_class, optimizer_class)
else:
raise RuntimeError("Wrong mode {} selected.".format(config.mode))
'triple_order': 'htr',
'delimiter': ' ',
'batch_size': 10000,
'num_workers': 10,
'num_evaluation_workers': 10,
'model': "TransE",
'optimizer': "SGD",
'margin': 1.0,
'epochs': 1000,
'lambda_': 0.001,
'evaluation_load_factor': 0.01
}
config = utils.build_config_with_dict(default_args)
print(config.__dict__)
triple_source = data.TripleSource(config.data_dir, config.triple_order,
config.triple_delimiter)
model_class = utils.load_class_from_module(config.model, 'kgegrok.models',
'kgegrok.text_models')
evaluator = evaluation.ParallelEvaluator(config, triple_source)
for changed_config in ParameterGrid(grid):
d = {}
d.update(config.__dict__)
changed_config[
'name'] = "TransE-FB15k237-neg_e_{}-neg_r_{}-ent_dim_{}-alpha_{}".format(
changed_config['negative_entity'],
changed_config['negative_relation'],
changed_config['entity_embedding_dimension'],
changed_config['alpha'])
d.update(changed_config)
if os.path.exists(
def source(config):
triple_dir = 'kgegrok/tests/fixtures/triples'
return data.TripleSource(triple_dir, 'hrt', ' ')