rnn.py 文件源码

def __call__(self, sess, epoch, iteration, model, loss):
        if iteration == 0 and epoch % self.at_every_epoch == 0:
            total = 0
            total_old = 0
            correct_old = 0
            correct = 0
            for values in self.batcher:
                total_old += len(values[-1])
                feed_dict = {}
                for i in range(0, len(self.placeholders)):
                    feed_dict[self.placeholders[i]] = values[i]
                truth = np.argmax(values[-1], 1)

                # mask truth
                truth_noneutral = ma.masked_values(truth, 0)
                truth_noneutral_compr = truth_noneutral.compressed()

                predicted = sess.run(tf.arg_max(tf.nn.softmax(model), 1),
                                     feed_dict=feed_dict)

                pred_nonneutral = ma.array(predicted, mask=truth_noneutral.mask)
                pred_nonneutral_compr = pred_nonneutral.compressed()

                correct_old += sum(truth == predicted)
                correct += sum(truth_noneutral_compr == pred_nonneutral_compr)
                total += len(truth_noneutral_compr)

            acc = float(correct) / total
            self.update_summary(sess, iteration, "AccurayNonNeut", acc)
            print("Epoch " + str(epoch) +
                  "\tAccNonNeut " + str(acc) +
                  "\tCorrect " + str(correct) + "\tTotal " + str(total))
            return acc
        return 0.0