CFDNN_MLP_03.py

for epoch in range(epochs):

    # Get our training input.
    user_input, item_input, labels = get_train_instances()

    # Generate a list of minibatches.
    minibatches = random_mini_batches(user_input, item_input, labels)

    # This has noting to do with tensorflow but gives
    # us a nice progress bar for the training
    progress = tqdm(total=len(minibatches))

    # Loop over each batch and feed our users, items and labels
    # into our graph. 
    for minibatch in minibatches:
        feed_dict = {user: np.array(minibatch[0]).reshape(-1,1),
                    item: np.array(minibatch[1]).reshape(-1,1),
                    label: np.array(minibatch[2]).reshape(-1,1)}
   
        # Execute the graph.
        _, l = session.run([step, loss], feed_dict)

        # Update the progress
        progress.update(1)
        progress.set_description('Epoch: %d - Loss: %.3f' % (epoch+1, l))

    progress.close()


# Calculate top@K    
hits = evaluate(df_neg)
print(np.array(hits).mean())