70888e622cf22e66188a74a0c29cf8d5f51ad9f9,basic/model.py,Model,_build_loss,#Model#,263

Before Change

        N, M, JX, JQ, VW, VC = \
            config.batch_size, config.max_num_sents, config.max_sent_size, \
            config.max_ques_size, config.word_vocab_size, config.char_vocab_size
        ce_loss = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(
            self.logits, tf.cast(tf.reshape(self.y, [-1, M * JX]), "float")))
        tf.add_to_collection("losses", ce_loss)
        ce_loss2 = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(
            self.logits2, tf.cast(tf.reshape(self.y2, [-1, M * JX]), "float")))
        tf.add_to_collection("losses", ce_loss2)

After Change

        N, M, JX, JQ, VW, VC = \
            config.batch_size, config.max_num_sents, config.max_sent_size, \
            config.max_ques_size, config.word_vocab_size, config.char_vocab_size
        loss_mask = tf.reduce_max(tf.cast(self.q_mask, "float"), 1)
        losses = tf.nn.softmax_cross_entropy_with_logits(
            self.logits, tf.cast(tf.reshape(self.y, [-1, M * JX]), "float"))
        ce_loss = tf.reduce_mean(loss_mask * losses)
        tf.add_to_collection("losses", ce_loss)
        ce_loss2 = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(
            self.logits2, tf.cast(tf.reshape(self.y2, [-1, M * JX]), "float")))
        tf.add_to_collection("losses", ce_loss2)

Instances