Tensorflow学习：Tensorboard可视化

Tensorflow学习：Tensorboard可视化

import tensorflow as tf
from tensorflow.examples.tutorials.mnist import input_data
from tensorflow.contrib.tensorboard.plugins import projector

# 载入数据集
mnist = input_data.read_data_sets("MNST_data", one_hot=True)
# 运行次数
max_steps = 1001
# 图片数量
image_num = 3000

# 文件路径
DIR = "D:/workspace/PyCharm/venv/cuiyongling"

# 定义会话
sess = tf.Session()

# 载入图片tf.stack():数据打包：手写数字测试集图片从0-3000张打一个包
embedding = tf.Variable(tf.stack(mnist.test.images[:image_num]), trainable=False, name="embedding")


# 定义一个函数：参数概要
def variable_summaries(var):
    with tf.name_scope('summaries'):
        mean = tf.reduce_mean(var)
        tf.summary.scalar('mean', mean)  # 平均值
        with tf.name_scope('stddev'):
            stddev = tf.sqrt(tf.reduce_mean(tf.square(var - mean)))
        tf.summary.scalar('stddev', stddev)  # 标准差
        tf.summary.scalar('max', tf.reduce_max(var))  # 最大值
        tf.summary.scalar('min', tf.reduce_min(var))  # 最小值
        tf.summary.histogram('histogram', var)  # 直方图


# 命名空间
with tf.name_scope("input"):
    # None:表示第一个维度可以使任意的张量
    x = tf.placeholder(tf.float32, [None, 784], name="x_input")
    # 正确的标签
    y = tf.placeholder(tf.float32, [None, 10], name="y_input")

# 显示图片
with tf.name_scope("input_reshape"):
    '''把x转换成[-1, 28, 28, 1]的形状：-1代表不确定的值，28,28：表示将784转化为28行28列
    原来是一个照片是一行784的张量。1：表示我们的维度，当图片是黑白的时候这里是1，当是彩色的时候这里是2'''
    image_reshaped_input = tf.reshape(x, [-1, 28, 28, 1])
    # 一次放10张图片
    tf.summary.image('input', image_reshaped_input, 10)

with tf.name_scope("layer"):
    # 创建一个简单神经网络
    with tf.name_scope("weights"):
        W = tf.Variable(tf.zeros([784, 10]), name="W")
        variable_summaries(W)
    with tf.name_scope("biases"):
        b = tf.Variable(tf.zeros([10]), name="b")
        variable_summaries(b)
    with tf.name_scope("wx_plus_b"):
        wx_plus_b = tf.matmul(x, W) + b
    with tf.name_scope("softmax"):
        prediction = tf.nn.softmax(wx_plus_b)

with tf.name_scope("loss"):
    # 交叉熵代价函数
    loss = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(labels=y, logits=prediction))
    tf.summary.scalar("loss", loss)
with tf.name_scope("train"):
    # 使用梯度下降法
    train_step = tf.train.GradientDescentOptimizer(0.5).minimize(loss)

# 初始化变量
sess.run(tf.global_variables_initializer())

with tf.name_scope("accuracy"):
    with tf.name_scope("correct_prediction"):
        # 结果存放在一个布尔型列表中:argmax返回一维张量中最大的值所在的位置，也就等于这个值
        correct_prediction = tf.equal(tf.argmax(y, 1), tf.argmax(prediction, 1))
    with tf.name_scope("accuracy"):
        # 求准确率:把correct_prediction布尔型的转变为float32类型
        accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))
        tf.summary.scalar("accuracy", accuracy)

# 产生metadata文件
if tf.io.gfile.exists(DIR + 'projector/projector/metadata.tsv'):
    # 存在的话删除
    tf.io.gfile.rmtree(DIR + 'projector/projector/metadata.tsv')
# 没有这个文件，以写的方式打开，会生成这个文件
with open(DIR + '/projector/projector/metadata.tsv', 'w') as f:
    # 得到测试集中的标签：使用one-hot编码，argmax返回一维张量中最大的值所在的位置，也就等于这个值（标签）
    labels = sess.run(tf.argmax(mnist.test.labels[:], 1))
    for i in range(image_num):
        # 把测试集的标签写入到文件中，写一行换行
        f.write(str(labels[i]) + '\n')

# 合并所有的summary
merged = tf.summary.merge_all()

# 根据Tensorboard官网例子进行修改
# 把图的结构存起来
projector_writer = tf.summary.FileWriter(DIR + 'projector/projector', sess.graph)
# 保存网络模型的
saver = tf.train.Saver()
# 定义配置项
config = projector.ProjectorConfig()
# 对配置项进行设置
embed = config.embeddings.add()
# 把embedding的名字赋给tensor的名字
embed.tensor_name = embedding.name
# 给他metadata的路径
embed_metadata_path = DIR + 'projector/projector/metadata.tsv'
# 给他图片的路径
embed.sprite.image_path = DIR + 'projector/data/mnist_10k_sprite.png'
# 对图片进行切分，按照28*28的像素进行切分
embed.sprite.single_image_dim.extend([28, 28])
# 嵌入可视化的一个工具
projector.visualize_embeddings(projector_writer, config)

for i in range(max_steps):
    # 每个批次100个样本
    batch_xs, batch_ys = mnist.train.next_batch(100)
    # 进行配置
    run_options = tf.RunOptions(trace_level=tf.RunOptions.FULL_TRACE)
    run_metadata = tf.RunMetadata()
    summary, _ = sess.run([merged, train_step], feed_dict={x: batch_xs, y: batch_ys}, options=run_options,
                          run_metadata=run_metadata)
    # 记录参数变化
    projector_writer.add_run_metadata(run_metadata, 'step%03d' % i)
    projector_writer.add_summary(summary, i)

    if i % 100 == 0:
        acc = sess.run(accuracy, feed_dict={x: mnist.test.images, y: mnist.test.labels})
        print("Iter " + str(i) + ",Testing Accuracy = " + str(acc))
# 把训练好的模型保存下来
saver.save(sess, DIR + 'projector/projector/a_model.ckpt', global_step=max_steps)
projector_writer.close()
sess.close()

输出文件：

cmd：metadata所在的目录

C:\Users\zhzsdiligence>d:

D:\>tensorboard --logdir=D:\workspace\PyCharm\venv\cuiyongling\projector\projector
d:\program files\python\lib\site-packages\numpy\_distributor_init.py:32: UserWarning: loaded more than 1 DLL from .libs:
d:\program files\python\lib\site-packages\numpy\.libs\libopenblas.NOIJJG62EMASZI6NYURL6JBKM4EVBGM7.gfortran-win_amd64.dll
d:\program files\python\lib\site-packages\numpy\.libs\libopenblas.PYQHXLVVQ7VESDPUVUADXEVJOBGHJPAY.gfortran-win_amd64.dll
  stacklevel=1)
2020-08-04 18:06:25.893724: W tensorflow/stream_executor/platform/default/dso_loader.cc:59] Could not load dynamic library 'cudart64_101.dll'; dlerror: cudart64_101.dll not found
2020-08-04 18:06:25.901551: I tensorflow/stream_executor/cuda/cudart_stub.cc:29] Ignore above cudart dlerror if you do not have a GPU set up on your machine.
Serving TensorBoard on localhost; to expose to the network, use a proxy or pass --bind_all
TensorBoard 2.3.0 at http://localhost:6006/ (Press CTRL+C to quit)

1、准确率：

2、偏置值和权重的最大值、最小值、平均值、标准差

3、损失代价函数：

4、训练集中的10张图片：

5、网络结构图：

6、偏置值和权值的分布图


附件：所需数据：链接: https://pan.baidu.com/s/1NVF8wUOpfvRb2KMFBLXHwg 提取码: df33