#number of input char types
char_vocab = len(string.printable)
# number of output classes = vocab size
numOutputClass = len(labelCorpus.dictionary)
print("Number of Classes:" + str(numOutputClass))
# Initialize models and start training
encoder = CharCNN(char_vocab, args.hidden_size)
decoder = DecoderRNN(args.hidden_size, numOutputClass)
encoder_optimizer = torch.optim.Adam(encoder.parameters(),
lr=args.learning_rate)
decoder_optimizer = torch.optim.Adam(decoder.parameters(),
lr=args.learning_rate)
criterion = nn.CrossEntropyLoss()
if args.cuda:
criterion.cuda()
encoder.cuda()
decoder.cuda()
start = time.time()
all_losses = []
loss_avg = 0
try:
dataloader = load_datasets(args)
best_error = 1000
early_stop = 0
if not os.path.exists(args.savedir):
os.makedirs(args.savedir)
model_name = args.savedir + '/' + 'best.pt'
train_begin = time.time()
print('train begin', '-' * 50)
print()
print()
model = CharCNN(70, args.dropout)
criterion = nn.CrossEntropyLoss()
if args.optimizer == 'SGD':
optimizer = optim.SGD(model.parameters(), lr=0.01, momentum=0.9)
scheduler = StepLR(optimizer, step_size=3, gamma=0.5)
elif args.optimizer == 'Adam':
optimizer = optim.Adam(model.parameters(), lr=0.0001)
else:
print('optimizer is bad')
optimizer = None
exit(0)
writer = SummaryWriter('log')
for epoch in range(args.epochs):
epoch_begin = time.time()
print('train {}/{} epoch'.format(epoch + 1, args.epochs))
train_loss = train_model(dataloader['train_dataloader'], model,
criterion, optimizer)