def train_drqa(cmdline_args, use_cuda=True):
# Parse cmdline args and setup environment
parser = argparse.ArgumentParser(
'DrQA Document Reader',
formatter_class=argparse.ArgumentDefaultsHelpFormatter)
add_train_args(parser)
config.add_model_args(parser)
args = parser.parse_args(cmdline_args)
set_defaults(args)
# Set cuda
# args.cuda = not args.no_cuda and torch.cuda.is_available()
args.cuda = use_cuda and torch.cuda.is_available()
if args.cuda:
logger.info('Using GPU')
torch.cuda.set_device(args.gpu)
else:
logger.info('Using CPU')
if args.use_bert_embeddings:
assert args.cuda, 'Must use CUDA if doing BERT training.'
# NOTE: The code below is commented out because it is done in smallfry utils.
# Set random state
# np.random.seed(args.random_seed)
# torch.manual_seed(args.random_seed)
# if args.cuda:
# torch.cuda.manual_seed(args.random_seed)
# Set logging
# logger.setLevel(logging.INFO)
# fmt = logging.Formatter('%(asctime)s: [ %(message)s ]',
# '%m/%d/%Y %I:%M:%S %p')
# console = logging.StreamHandler()
# console.setFormatter(fmt)
# logger.addHandler(console)
# if args.log_file:
# if args.checkpoint:
# logfile = logging.FileHandler(args.log_file, 'a')
# else:
# logfile = logging.FileHandler(args.log_file, 'w')
# logfile.setFormatter(fmt)
# logger.addHandler(logfile)
# logger.info('COMMAND: %s' % ' '.join(sys.argv))
# Run!
return main(args)
pin_memory=args.cuda,
)
stats = {'timer': utils.Timer(), 'epoch': 0, 'best_valid': 0}
scores, pred_objs = compute_expected_metric(args, dev_loader, model, stats,
dev_offsets, dev_texts,
dev_answers)
return scores, pred_objs
if __name__ == '__main__':
# Parse cmdline args and setup environment
parser = argparse.ArgumentParser(
'DrQA Document Reader',
formatter_class=argparse.ArgumentDefaultsHelpFormatter)
add_train_args(parser)
config.add_model_args(parser)
args = parser.parse_args()
set_defaults(args)
# Set cuda
args.cuda = not args.no_cuda and torch.cuda.is_available()
if args.cuda:
torch.cuda.set_device(args.gpu)
# Set random state
np.random.seed(args.random_seed)
torch.manual_seed(args.random_seed)
if args.cuda:
torch.cuda.manual_seed(args.random_seed)
# Set logging
if result[args.valid_metric] > stats['best_valid']:
logger.info('Best valid: %s = %.2f (epoch %d, %d updates)' %
(args.valid_metric, result[args.valid_metric],
stats['epoch'], model.updates))
model.save(args.model_file)
stats['best_valid'] = result[args.valid_metric]
if __name__ == '__main__':
# Parse cmdline args and setup environment
parser = argparse.ArgumentParser(
'DrQA Document Reader',
formatter_class=argparse.ArgumentDefaultsHelpFormatter
)
add_train_args(parser)
config.add_model_args(parser)
args = parser.parse_args()
set_defaults(args)
# Set cuda
args.cuda = not args.no_cuda and torch.cuda.is_available()
if args.cuda:
torch.cuda.set_device(args.gpu)
# Set random state
np.random.seed(args.random_seed)
torch.manual_seed(args.random_seed)
if args.cuda:
torch.cuda.manual_seed(args.random_seed)
# Set logging
action='store_true',
help='Only store single top span instead of top N list')
parser.add_argument('--use-default', action="store_true")
args = parser.parse_args()
t0 = time.time()
args.cuda = not args.no_cuda and torch.cuda.is_available()
if args.cuda:
torch.cuda.set_device(args.gpu)
logger.info('CUDA enabled (GPU %d)' % args.gpu)
else:
logger.info('Running on CPU only.')
if args.use_default:
other_parser = argparse.ArgumentParser()
add_model_args(other_parser)
other_args = other_parser.parse_args([])
else:
print("no other args")
other_args = None
predictor = Predictor(model=args.model,
tokenizer=args.tokenizer,
embedding_file=args.embedding_file,
num_workers=args.num_workers,
other_args=other_args)
if args.cuda:
predictor.cuda()
# ------------------------------------------------------------------------------
# Read in dataset and make predictions.
