在使用 TensorFlow 进行多变量线性回归的情况下，如何添加 CSV 日志记录机制？

Question

假设，下面是我在Python中的多元线性回归源代码：

import os

os.environ["TF_CPP_MIN_LOG_LEVEL"] = "2"

import sys, random
import time

import tensorflow as tf
from   tensorflow import keras
from   tensorflow.keras.models import Sequential
from   tensorflow.keras.layers import Dense
from   tensorflow.keras.optimizers import Adam
import matplotlib.pyplot as plt
import numpy as np

def load_data_k(fname: str, yyy_index: int, **selection):
    i = 0
    file = open(fname)
    if "top_n_lines" in selection:
        lines = [next(file) for _ in range(int(selection["top_n_lines"]))]
    elif "random_n_lines" in selection:
        tmp_lines = file.readlines()
        lines = random.sample(tmp_lines, int(selection["random_n_lines"]))
    else:
        lines = file.readlines()

    data_x, data_y = [], []
    for l in lines:
        row = l.strip().split()   
        x = [float(ix) for ix in row[yyy_index+1:]] 
        y = float(row[yyy_index])  
        data_x.append(x) 
        data_y.append(y) 
    # END for l in lines...

    num_rows = len(data_x)

    print("row size = ", len(data_x[0]))

    given_fraction = selection.get("validation_part", 1.0)
    if given_fraction > 0.9999:
        valid_x, valid_y = data_x, data_y
    else:
        n = int(num_rows * given_fraction)
        data_x, data_y = data_x[n:], data_y[n:]
        valid_x, valid_y = data_x[:n], data_y[:n]
    # END of if-else block

    print("size of x = ", len(data_x))
    print("size of y = ", len(data_y))

    tx = tf.convert_to_tensor(data_x, dtype=tf.float32)
    ty = tf.convert_to_tensor(data_y, dtype=tf.float32)

    vx = tf.convert_to_tensor(valid_x, dtype=tf.float32)
    vy = tf.convert_to_tensor(valid_y, dtype=tf.float32)

    return tx, ty, vx, vy
# END of the function


# load training data from the disk
train_x, train_y, validate_x, validate_y = \
        load_data_k(
            fname="data_file.csv",
            yyy_index=6,
            random_n_lines=90000,
            validation_part=0.2
        )

print("training data size : ", len(train_x))
print("validation data size : ", len(validate_x))

predict_data = np.array([[7.042, 5.781, 5.399, 5.373, 5.423, -9.118, 5.488,   5.166,   4.852,  7.470,  6.452,  6.069,
                         0,     0,     0,     1,     0,      1,     1,       3,       0,      0,      0,      1, 
                         0,     0,     0,     0,     0,      0,     0,       0,       0,      0,      0,      0, 
                         0,     0,     0,     0]])


# Create Keras model
model = Sequential()
model.add(Dense(1, input_dim=40))
model.add(Dense(128))
model.add(Dense(128))
model.add(Dense(1))

# Gradient descent algorithm
adam_opt = Adam(0.1)

model.compile(loss='mse', optimizer=adam_opt)

history = model.fit(train_x, train_y, epochs=500)

prediction = model.predict(predict_data)
print(prediction)

我想为训练损失、验证损失、训练准确性和验证准确性添加 CSV 日志记录。

我该怎么做？

Answer 1

只需使用 tf.keras.callbacks.CSVLogger 和任何您想在训练期间记录的回归指标：

import tensorflow as tf

model = tf.keras.Sequential()
model.add(tf.keras.layers.Dense(1, input_dim=40))
model.add(tf.keras.layers.Dense(128))
model.add(tf.keras.layers.Dense(128))
model.add(tf.keras.layers.Dense(1))

adam_opt = tf.keras.optimizers.Adam(0.1)
model.compile(loss='mse', optimizer=adam_opt, metrics=tf.keras.metrics.MeanSquaredError(name="mean_squared_error", dtype=None))

train_x = tf.random.normal((50, 40))
train_y = tf.random.normal((50, 1))
val_x = tf.random.normal((50, 40))
val_y = tf.random.normal((50, 1))

csv_logger = tf.keras.callbacks.CSVLogger('model_training.csv')
history = model.fit(train_x, train_y, epochs=5, validation_data=(val_x, val_y), callbacks=[csv_logger])

model_training.csv:

epoch        loss  mean_squared_error    val_loss  val_mean_squared_error
    0  304.349060          304.349060   69.584991               69.584991
    1  105.304787          105.304787  170.063126              170.063126
    2  175.232788          175.232788    7.874812                7.874812
    3  104.159607          104.159607  320.626556              320.626556
    4  194.709763          194.709763    1.438866                1.438866

在使用 TensorFlow 进行多变量线性回归的情况下，如何添加 CSV 日志记录机制？

How can I add CSV logging mechanism in case of Multivariable Linear Regression using TensorFlow?

python

logging

deep-learning

keras

tensorflow