ModelRunner.py

#Build the model and add to graphics card
model = ModelRNNBasic().cuda()

# Define Loss, Optimizer
lr=1e-3
opt = torch.optim.Adam(model.parameters(), lr=lr)
def loss_func(input,target):
  return F.cross_entropy(input.permute(0,2,1), target)

losses = []
accuracy = []

#each epoch
n_epochs = 10
for epoch in range(1,n_epochs+1):

  data = DataLM(train_list,bptt=70)

  #each batch
  for x,y in data:
    #opt = torch.optim.Adam(model.parameters(), lr)
    y_hat = model(x)
    loss = loss_func(y_hat, y)
    loss.backward()
    opt.step()
    opt.zero_grad()

    _, predicted = torch.max(y_hat[:,-1,:], 1)
    
    correct = (predicted == y[:,-1]).sum().item()

    losses.append(loss.item())
    accuracy.append(correct / 64)
  print('Epoch: {:2}/{}   |   Loss: {:.4f}   |   Accuracy: {:.4f}'.format(epoch, n_epochs,np.mean(losses[-3:]),np.mean(accuracy[-3:])))