How to ensure the execution of TF model the same every time?

[shaoyu_young]

Hi, guys.
I am a newcomer to TensorFlow.
For the following code, I found that the result is different every time I execute it.

I have used many ways (e.g., tf.random.set_seed, os.environ[‘TF_DETERMINISTIC_OPS’] = ‘1’) but all of these settings do not work.
really thanks if anyone can provide some helps

[Kiran_Sai_Ramineni]

@shaoyu_young

For the following code, I found that the result is different every time I execute it.

Hi @shaoyu_young, when defining the simple RNN layer set the seed to the kernel, recurrent initializer to get the same result. please refer to this gist for working code example. Thank You.

[shaoyu_young]

hi @Kiran_Sai_Ramineni , thank you very much! this helps me a lot.
If there is a list to tell me which APIs are similar to the RNN, it would be great!

[Kiran_Sai_Ramineni]

Hi @shaoyu_young, The layers similar to RNN are LSTM, GRU which have some advanced features. Thank You.

[shaoyu_young]

thank you, @Kiran_Sai_Ramineni .
Actually, I mean which operators are similar to RNN which needs to be manually initialized, avoiding random results.