cnn.keep_prob: 1.
})
#print loss
ou = sess.run(pred,
feed_dict={
cnn.x: batch_x,
cnn.y: batch_y,
cnn.keep_prob: 1
})
#print ou.shape
#print batch_y.shape
[
hammin_loss, one_error, coverage, ranking_loss,
average_precision, subset_accuracy, accuracy, precision,
recall, f_beta
] = utils.get_accuracy_test(ou, batch_y)
#print(acc)
plot_x.append(step * config.batch_size)
plot_y.append(loss)
print("Iter " + str(step * config.batch_size) +
", Minibatch Loss= " + "{:.6f}".format(loss[0]))
print("hammin_loss: ", "{:.6f}".format(hammin_loss))
print("subset_accuracy: ", "{:.6f}".format(subset_accuracy))
print("accuracy: ", "{:.6f}".format(accuracy))
print("precision: ", "{:.6f}".format(precision))
print("recall: ", "{:.6f}".format(recall))
print("f_beta: ", "{:.6f}".format(f_beta))
if data.end == data.total_texts:
epoch += 1
print("Epoch: " + str(epoch))
data.shuffler()
#batch_y = data.labels_train # bibtex, rcv1
batch_y = np.array(data.labels_train) # agnews
batch_y = batch_y.reshape(config.batch_size, config.label_size)
#print(len(batch_x), len(batch_x[0]))
#print("Y shape: ", batch_y.shape)
sess.run(optimizer, feed_dict={mlp.x: batch_x, mlp.y: batch_y, mlp.keep_prob: mlp.dropout})
if step % 1 == 0:
#print Get Accuracy: "
loss = sess.run([cost], feed_dict={mlp.x: batch_x, mlp.y: batch_y, mlp.keep_prob: 1.})
#print loss
ou = sess.run(pred, feed_dict={mlp.x: batch_x, mlp.y: batch_y, mlp.keep_prob: 1})
#data.ids[data.start:data.end]
[hammin_loss, one_error, coverage, ranking_loss, average_precision, subset_accuracy, accuracy, precision, recall, f_beta] = utils.get_accuracy_test(ou, batch_y)
#print(data.ids[data.start:data.end])
print ("Iter " + str(step * config.batch_size) + ", Minibatch Loss= " + "{:.6f}".format(loss[0]))
print ("hammin_loss: ", "{:.6f}".format(hammin_loss))
print ("subset_accuracy: ", "{:.6f}".format(subset_accuracy))
print ("accuracy: ", "{:.6f}".format(accuracy))
print ("precision: ", "{:.6f}".format(precision))
print ("recall: ", "{:.6f}".format(recall))
print ("f_beta: ", "{:.6f}".format(f_beta))
if data.end == data.total_texts:
epoch += 1
print("Epoch: " + str(epoch))
data.shuffler()
if step % 5000 == 0:
save_path = saver.save(sess, "mlp_weights/model" + str(model_saving) + "_bow.ckpt")