Stable results with TensorFlow

What would be a good way to ensure that TensorFlow results keep different runs? I can not find much information, for example. initialization of a random sowing weight, so my results tend to vary with each run.

+5
source share
1 answer

The tf.set_random_seed() API method can be used to set a random seed that will be used in all TensorFlow random operations (including regular random weight initializers and tf.RandomShuffleQueue ).

+7
source

Source: https://habr.com/ru/post/1238198/


All Articles