CODEWITHSUNDEEP
tensorflowneural-networkdeep-learning
Master Yii in Training
Master Yii in Training

Reputation: 95

Is there any way to verify that I have successfully frozen certain variables/parameters in my Tensorflow model?

I attempted to freeze some of the variables by only passing the variables that I would like to fine-tune on via the

 tf.train.AdamOptimizer.compute_gradients(var_list=[])

approach.

But when I proceeded with fine-tuning my model, the loss is going up and the accuracy is going down so I am having doubt that I have successfully frozen the initial few layers.

Any way to check if I'm freezing variables properly? Thanks!

Upvotes: 2

Views: 341

Answers (1)

from keras import michael
from keras import michael

Reputation: 721

Try tf.get_collection(tf.GraphKeys.TRAINABLE_VARIABLES).

Upvotes: 1

Related Questions