loss = criterion(output, y.view(-1).long())
loss.backward()
optimizer.step()
if epoch % 10 == 0:
print("Epoch: {}/{}.............".format(epoch, n_epochs), end=" ")
print("Loss: {:.4f}".format(loss.item()))
def predict(model, question):
"""
Returns the answer to the question.
"""
question = question.ljust(dataset.longer_question_length)
question = dataset.text2int(question)
question = dataset.one_hot_encode(question)
question = torch.from_numpy(np.array([question])).float().cuda()
output, hidden = model(question)
answer = dataset.one_hot_decode(output.cpu())
answer = dataset.int2text(answer)
return answer
model.eval()
with torch.no_grad():
prediction = predict(model, "how are yo?")
print("".join(prediction))
output = net(x_train)
del x_train
batch_loss = criterion(output, targets.contiguous().view(-1))
del targets
batch_loss.backward()
loss += float(batch_loss.item())
optimizer.step()
del batch_loss
j += 1
loss = loss / j
net.init_hidden()
net.eval()
val_loss = 0.
j = 0
for val_x, val_y in get_batches(val_data, sequence_len, batch_size):
val_x = torch.from_numpy(val_x).type(torch.LongTensor).to(net.device)
val_y = torch.from_numpy(val_y.T).type(
torch.LongTensor).contiguous().view(-1).to(net.device)
val_output = net(val_x)
#val_output = net(val_x)
batch_val_loss = float(
criterion(val_output,
val_y.contiguous().view(-1)).item())
val_loss += batch_val_loss
del batch_val_loss
j += 1
