必须设置哪些种子才能在 tensorflow 中实现 100% 的训练结果可重复性? [英] Which seeds have to be set where to realize 100% reproducibility of training results in tensorflow?

问题描述

在一般的 tensorflow 设置中，如

model =construct_model()使用 tf.Session() 作为 sess:train_model(sess)

其中 construct_model() 包含模型定义，包括权重的随机初始化 (tf.truncated_normal) 和 train_model(sess) 执行训练模型的-

在重复运行上述代码片段之间，我必须设置哪些种子以确保 100% 的可重复性?文档 tf.random.set_random_seed 可能简洁，但让我有点困惑.我试过了:

tf.set_random_seed(1234)模型 = 构造模型()使用 tf.Session() 作为 sess:train_model(sess)

但每次都得到不同的结果.

解决方案

目前适用于 GPU 的最佳解决方案是使用以下内容安装 tensorflow-determinism:

pip install tensorflow-determinism

然后在您的代码中包含以下代码

 将 tensorflow 导入为 tf导入操作系统os.environ['TF_DETERMINISTIC_OPS'] = '1'

来源:https://github.com/NVIDIA/tensorflow-determinism>

In a general tensorflow setup like

model = construct_model()
with tf.Session() as sess:
    train_model(sess)

Where construct_model() contains the model definition including random initialization of weights (tf.truncated_normal) and train_model(sess) executes the training of the model -

Which seeds do I have to set where to ensure 100% reproducibility between repeated runs of the code snippet above? The documentation for tf.random.set_random_seed may be concise, but left me a bit confused. I tried:

tf.set_random_seed(1234)
model = construct_model()
    with tf.Session() as sess:
        train_model(sess)

But got different results each time.

解决方案

The best solution which works as of today with GPU is to install tensorflow-determinism with the following:

pip install tensorflow-determinism

Then include the following code to your code

import tensorflow as tf
import os
os.environ['TF_DETERMINISTIC_OPS'] = '1'

source: https://github.com/NVIDIA/tensorflow-determinism

这篇关于必须设置哪些种子才能在 tensorflow 中实现 100% 的训练结果可重复性?的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！