def main():
embedding = np.load("data/wordvec_200.npy")
vocab_size = embedding.shape[0]
model = MwAN(embedding, vocab_size=vocab_size, embedding_size=args.emsize, encoder_size=args.nhid, drop_out=args.dropout)
logger.info(model)
logger.info("path to save the model {}".format(args.save))
print('Model total parameters:', get_model_parameters(model))
if args.cuda:
model.cuda()
if args.mode == 'train':
optimizer = torch.optim.Adamax(model.parameters())
else:#'continue'
optimizer = torch.optim.Adam(model.parameters(), lr=1e-4, betas=(0.9, 0.999), eps=1e-8, weight_decay=0)
with open(args.data + 'train.pickle', 'rb') as f:
train_data = pickle.load(f)
with open(args.data + 'dev.pickle', 'rb') as f:
dev_data = pickle.load(f)
dev_data = sorted(dev_data, key=lambda x: len(x[1]))
logger.info('train data size {:d}, dev data size {:d}'.format(len(train_data), len(dev_data)))
best = 0.0
for epoch in range(args.epoch):
acc = train(epoch, best, model, optimizer, train_data, dev_data)
if acc > best:
best = acc
acc = test(model, dev_data)
if acc > best:
best = acc
with open(args.save, 'wb') as f:
torch.save(model, f)
f.close()
logger.info('epcoh {:d} validation acc is {:f}, best validation acc {:f}'.format(epoch, acc, best))
type=str,
default='model.pt',
help='path to save the final model')
parser.add_argument('--word_path',
type=str,
default='data/word2id.obj',
help='location of the word2id.obj')
args = parser.parse_args()
# vocab_size = process_data(args.data, args.threshold)
vocab_size = 98745
model = MwAN(vocab_size=vocab_size,
embedding_size=args.emsize,
encoder_size=args.nhid,
drop_out=args.dropout)
print('Model total parameters:', get_model_parameters(model))
print(model)
if args.cuda:
model.cuda()
optimizer = torch.optim.Adamax(model.parameters())
with open(args.data + 'train.pickle', 'rb') as f:
train_data = cPickle.load(f)
with open(args.data + 'dev.pickle', 'rb') as f:
dev_data = cPickle.load(f)
dev_data = sorted(dev_data, key=lambda x: len(x[1]))
print('train data size {:d}, dev data size {:d}'.format(
len(train_data), len(dev_data)))
help='# of batches to see the training error')
parser.add_argument('--dropout', type=float, default=0.1,
help='dropout applied to layers (0 = no dropout)')
parser.add_argument('--cuda', action='store_true',
help='use CUDA')
parser.add_argument('--save', type=str, default='model_1.pt',
help='path to save the final model')
args = parser.parse_args()
# vocab_size = process_data(args.data, args.threshold)
embedding = np.load("/home/zhengyinhe/xie_data/data/tencent_wordvec_200.npy")
vocab_size = embedding.shape[0]
print(vocab_size)
model = MwAN(embedding, vocab_size, heads=8, embedding_size=args.emsize,
encoder_size=args.nhid, conv_num=4, attn_num=1, drop_out=args.dropout)
logger.info(model)
logger.info("path to save the model {}".format(args.save))
print('Model total parameters:', get_model_parameters(model))
if args.cuda:
model.cuda()
optimizer = torch.optim.Adamax(model.parameters())
with open(args.data + 'train.pickle', 'rb') as f:
train_data = pickle.load(f)
with open(args.data + 'dev.pickle', 'rb') as f:
dev_data = pickle.load(f)
dev_data = sorted(dev_data, key=lambda x: len(x[1]))
logger.info('train data size {:d}, dev data size {:d}'.format(len(train_data), len(dev_data)))
help='of batches to see the validation result')
parser.add_argument('--dropout', type=float, default=0.0,
help='dropout applied to layers (0 = no dropout)')
parser.add_argument('--cuda', action='store_true',
help='use CUDA')
parser.add_argument('--save', type=str, default='model_3.pt',
help='path to save the final model')
args = parser.parse_args()
# vocab_size = process_data(args.data, args.threshold)
embedding = np.load("/home/zhengyinhe/xie_data/data2/tencent_wordvec_200.npy")
vocab_size = embedding.shape[0]
print(vocab_size)
model = MwAN(embedding, vocab_size=vocab_size, embedding_size=args.emsize, encoder_size=args.nhid, drop_out=args.dropout)
logger.info(model)
logger.info("path to save the model {}".format(args.save))
print('Model total parameters:', get_model_parameters(model))
if args.cuda:
model.cuda()
embedding_params = list(map(id, model.embedding.parameters()))
base_params = filter(lambda p: id(p) not in embedding_params,
model.parameters())
with open(args.data + 'train.pickle', 'rb') as f:
train_data = pickle.load(f)
with open(args.data + 'dev.pickle', 'rb') as f:
dev_data = pickle.load(f)
dev_data = sorted(dev_data, key=lambda x: len(x[1]))