TensorBoard是通過讀取tensorflow產(chǎn)生的事件文檔來運(yùn)行的．這些文檔中包含著代碼運(yùn)行過程中產(chǎn)生的總結(jié)信息．下面本文將會(huì)對tensorboard進(jìn)行詳細(xì)的介紹．

１.生命周期

??首先，創(chuàng)建想要收集數(shù)據(jù)的tensorflow graph,然后指明想要收集數(shù)據(jù)的節(jié)點(diǎn)．summary是一個(gè)operation.收集操作.

??例如，假設(shè)你在ＭNIST數(shù)據(jù)集上訓(xùn)練卷積神經(jīng)網(wǎng)絡(luò)．你呢，想記錄隨著學(xué)習(xí)的深入，學(xué)習(xí)率的變化程度喲，還有目標(biāo)函數(shù)的變化喲．想要收集這些信息呢，就需要使用 tf.summary.scalar ．然后呢就要給這些總結(jié)信息 scalar_summary一個(gè)相對比較有意義的標(biāo)簽 tag,例如 'learning rate' ，'loss function'.等等

??也許呢，你還想可視化的觀察一下某一層的激活函數(shù)的輸出情況，或者是梯度和權(quán)重的分布情況．我們可以通過 tf.summary.histogram 來收集這些信息．想要對這些總總結(jié)信息有更深入的了解，就需要查看一下文檔． summary operations.

??Tensorflow中的operation在運(yùn)行之前什么都不會(huì)做．當(dāng)然，如果你運(yùn)行這個(gè)operation的一個(gè)下行operation也是可以觸發(fā)這個(gè)operation的運(yùn)行的．Summary的節(jié)點(diǎn)是外圍節(jié)點(diǎn)，沒有任何的操作節(jié)點(diǎn)是需要依賴這些節(jié)點(diǎn)的．因此，我們需要把這些總結(jié)節(jié)點(diǎn)一個(gè)個(gè)的自己運(yùn)行完，這很煩呀，所以有個(gè)快捷鍵．tf.summary.merge_all 可以將這些節(jié)點(diǎn)合并成一個(gè)節(jié)點(diǎn)，然后運(yùn)行這個(gè)節(jié)點(diǎn)就好啦．具體運(yùn)行的方法是使用writer.add_summary觸發(fā)．觸發(fā)一次就可以寫入這一批執(zhí)行的信息，不多不少．然后每次運(yùn)行的時(shí)候都會(huì)產(chǎn)生輸出信息，把這個(gè)節(jié)點(diǎn)傳給tf.summary.FileWriter就可以寫入到文件當(dāng)中．

?? FileWriter 使用logdir作為輸入?yún)?shù)．這個(gè)參數(shù)可以說是非常重要的．所有的輸出信息都會(huì)在這個(gè)目錄中輸出．同時(shí)，還可以選擇graph作為參數(shù)．通過接收graph，tensorboard就可以將整個(gè)圖像進(jìn)行可視化的展示啦． Tensor shape information.

??現(xiàn)在，你就可以修改你的graph增加一個(gè)FileWriter了,如果你愿意的話，你就可以在每一步都進(jìn)行一下summary 的操作，然后把每一步的信息都顯示出來．不過最好呢，是每隔n步再進(jìn)行一次信息的總結(jié)．下面是 simple MNIST tutorial的一個(gè)簡單的改編版本的代碼,在代碼中我們增加了一些總結(jié)的操作，每隔10步就會(huì)執(zhí)行一次．運(yùn)行tensorboard --logdir=/tmp/tensorflow/mnist, 你就可以看到可視化的數(shù)據(jù)．源代碼在這里.

def variable_summaries(var):
  """Attach a lot of summaries to a Tensor (for TensorBoard visualization)."""
  with tf.name_scope('summaries'):
    mean = tf.reduce_mean(var)
    tf.summary.scalar('mean', mean)
    with tf.name_scope('stddev'):
      stddev = tf.sqrt(tf.reduce_mean(tf.square(var - mean)))
    tf.summary.scalar('stddev', stddev)
    tf.summary.scalar('max', tf.reduce_max(var))
    tf.summary.scalar('min', tf.reduce_min(var))
    tf.summary.histogram('histogram', var)

def nn_layer(input_tensor, input_dim, output_dim, layer_name, act=tf.nn.relu):
  """Reusable code for making a simple neural net layer.

  It does a matrix multiply, bias add, and then uses relu to nonlinearize.
  It also sets up name scoping so that the resultant graph is easy to read,
  and adds a number of summary ops.
  """
  # Adding a name scope ensures logical grouping of the layers in the graph.
  with tf.name_scope(layer_name):
    # This Variable will hold the state of the weights for the layer
    with tf.name_scope('weights'):
      weights = weight_variable([input_dim, output_dim])
      variable_summaries(weights)
    with tf.name_scope('biases'):
      biases = bias_variable([output_dim])
      variable_summaries(biases)
    with tf.name_scope('Wx_plus_b'):
      preactivate = tf.matmul(input_tensor, weights) + biases
      tf.summary.histogram('pre_activations', preactivate)
    activations = act(preactivate, name='activation')
    tf.summary.histogram('activations', activations)
    return activations

hidden1 = nn_layer(x, 784, 500, 'layer1')

with tf.name_scope('dropout'):
  keep_prob = tf.placeholder(tf.float32)
  tf.summary.scalar('dropout_keep_probability', keep_prob)
  dropped = tf.nn.dropout(hidden1, keep_prob)

# Do not apply softmax activation yet, see below.
y = nn_layer(dropped, 500, 10, 'layer2', act=tf.identity)

with tf.name_scope('cross_entropy'):
  # The raw formulation of cross-entropy,
  #
  # tf.reduce_mean(-tf.reduce_sum(y_ * tf.log(tf.softmax(y)),
  #                               reduction_indices=[1]))
  #
  # can be numerically unstable.
  #
  # So here we use tf.losses.sparse_softmax_cross_entropy on the
  # raw logit outputs of the nn_layer above.
  with tf.name_scope('total'):
    cross_entropy = tf.losses.sparse_softmax_cross_entropy(labels=y_, logits=y)
tf.summary.scalar('cross_entropy', cross_entropy)

with tf.name_scope('train'):
  train_step = tf.train.AdamOptimizer(FLAGS.learning_rate).minimize(
      cross_entropy)

with tf.name_scope('accuracy'):
  with tf.name_scope('correct_prediction'):
    correct_prediction = tf.equal(tf.argmax(y, 1), tf.argmax(y_, 1))
  with tf.name_scope('accuracy'):
    accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))
tf.summary.scalar('accuracy', accuracy)

# Merge all the summaries and write them out to /tmp/mnist_logs (by default)
merged = tf.summary.merge_all()
train_writer = tf.summary.FileWriter(FLAGS.summaries_dir + '/train',
                                      sess.graph)
test_writer = tf.summary.FileWriter(FLAGS.summaries_dir + '/test')
tf.global_variables_initializer().run()

&esmp;?剩下的部分需要介紹的就不是很多啦．

&esmp;?&esmp;?1.name_scope對變量進(jìn)行分類有助于在tensorboard中進(jìn)行查看

&esmp;?&esmp;?2.session 是可以保存和恢復(fù)的

&esmp;?&esmp;?3.meta_data是運(yùn)行時(shí)的統(tǒng)計(jì)信息，包括運(yùn)行內(nèi)存啦什么的．