tensorboard 可视化模型结构图 探索

1. 实验

"""
test tensorboard basic demo
"""

from keras.layers import Dense
from keras.models import Sequential
from keras.callbacks import TensorBoard
import numpy as np

model = Sequential()

model.add(Dense(units=64, activation='relu', input_shape=(3,), name="dense1"))
model.add(Dense(units=1, name="dense2"))

model.compile(loss='binary_crossentropy',
              optimizer='sgd',
              metrics=['accuracy'])
x_train = np.array([[1, 2, 3], [4, 5, 6]])
y_train = np.array([1, 0])

model.fit(x_train, y_train, epochs=5, batch_size=32,
          callbacks=[
              TensorBoard(
                  log_dir="/tmp/test_tensorboard_dashboard", write_graph=True)]
          )
# 最终通过 tensorboard --logdir=/tmp/test_tensorboard_dashboard 进行查看。

2. tensorboard的具体结构图

权重随机初始化

赋值操作

bias赋值操作

metrics中准确率的具体计算

logloss的具体计算

损失求参数求梯度，并对相应参数进行更新和赋值