loss_cache = 10
for epoch in xrange(1, total_epoch + 1):
print 'Epoch: {}/{}'.format(epoch, total_epoch)
# Training (Mini-batch)
now = time.time()
data.shuffle()
bar = Bar(max_value=n)
bar.cursor.clear_lines(2) # Make some room
bar.cursor.save() # Mark starting line
for _ in xrange(data.batch_run()):
net.input(data.next_batch())
net.forward()
net.backward(lr, beta_1, beta_2, epoch)
bar.cursor.restore() # Return cursor to start
bar.draw(value=data.get_count())
t = time.time() - now
acc, loss = net.get_record()
loss_avg = np.array(loss).mean()
loss_diff = loss_avg - loss_cache
loss_cache = loss_avg
print 'Acc: ', np.array(acc).mean()
print 'Loss: ', loss_avg
print 'Time: ', t
f, b = net.get_profile()
net.clear_record()
bar_t = Bar(max_value=nt)
bar_t.cursor.clear_lines(2) # Make some room
bar_t.cursor.save() # Mark starting line
for _ in xrange(data.batch_run_test()):
