python - Recover Tensorflow variables from model

Question

I have been working with a TF model and I want to restore the model variables. However, these variables were never saved during the Training stage, and I'm wondering if there is a way to recover those from just the model I recreated from just the .h5 weights file.

This is how I'm recovering the model:

import tensorflow as tf

with tf.Session(config=conf) as sess:
    restored_model = my_custom_model(sess, ...)
    restored_model.load_weights('/path/to/file.h5')

The my_custom_model() definition initializes variables like this:

    # Init vars before training
    sess.run(tf.local_variables_initializer())
    sess.run(tf.global_variables_initializer())
    sess.run(tf.tables_initializer())
    K.set_session(sess)

However, during training, the model variables were not saved. Is there a way to get these variables?

Answer 1

等待大神答复