How do you put random seed in Tensorflow?

[guilherme]

I'm always getting different results with the same parameters. I need to get the same results if I didn't change the parameters.

Sold this code:

tensorflow.random.set_seed(4)

• It doesn't help, learning results are always different.

Neurosets:

from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import LSTM, Dense, Dropout, Flatten
from tensorflow.keras.metrics import AUC
from tensorflow_addons.metrics import F1Score
batch_size = 4

num_neurons = 100
n_model = Sequential()

n_model.add(LSTM(num_neurons, return_sequences=True, input_shape=(maxlen, emb_dims)))

n_model.add(Dropout(.2))

n_model.add(Dropout(.2))

n_model.add(Flatten())

n_model.add(Dense(n_classes, activation="softmax"))

n_model.compile('rmsprop', loss='categorical_crossentropy', metrics=['accuracy', F1Score(num_classes=15)])

[derk]

Could be through numpy. numpy.random.seed(seed)