In other words, is it safe to create an Estimator object on the main thread, and then call predict on childThread1 and train on childThread2? The purpose of doing so is to perform predictions with the latest trained model. Is this possible or will the simultaneous training & prediction pollute the internal state?
Asked
Active
Viewed 188 times
1
-
It should be thread-safe, see [Is it thread-safe when using tf.Session in inference service?](https://stackoverflow.com/q/38694111). – jdehesa Mar 13 '19 at 09:09