如何在张量流中用估计量绘制评估损失?

时间:2018-08-29 17:21:01

标签: tensorflow deep-learning conv-neural-network tensorboard

嗨,我想画出像训练损失这样的评估损失: image ,而不仅仅是tensorflow教程之类的点,我是怎么做的,这是我的代码,使用此代码,我只会得到一个点,表示评估损失:

accuracy=tf.metrics.accuracy(labels=labels, predictions=predictions["classes"])
    metrics = {"accuracy": accuracy}
    tf.summary.scalar("accuracy", accuracy[1])

    #Configure of the training operation
    if mode==tf.estimator.ModeKeys.TRAIN:
        optimizer = tf.train.GradientDescentOptimizer(learning_rate=0.001)
        train_op=optimizer.minimize(loss=loss,global_step=tf.train.get_global_step())
        return tf.estimator.EstimatorSpec(mode=mode,loss=loss,train_op=train_op)


    #Configure the evaluation operation
    if mode == tf.estimator.ModeKeys.EVAL:
        return tf.estimator.EstimatorSpec(mode=mode, loss=loss, eval_metric_ops=metrics)

当我在指标中输入“ loss”:损耗时,我会出现错误,该怎么办?

1 个答案:

答案 0 :(得分:0)

您必须指定要使用的损失类型:

查看文档:{​​{3}}

例如:loss='mean_squared_error'

相关问题