for epoch in range(1, 100):
    loss = 0.
    i = 0
    X,Y = generate_context_word_pairs(corpus=wids, window_size=window_size, vocab_size=vocab_size)
    for x, y in zip(X,Y):
        i += 1
        optimizer.zero_grad()
        log_probs = model(x[0])
        loss = loss_function(log_probs,torch.Tensor([y]).long())
        loss.backward()
        optimizer.step()
        loss += loss.data
    print('Epoch:', epoch, '\tLoss:', loss)