Tensorflow Embeddings Don't Exist After First Rnn Example

I've setup a print statement and I've noticed that for the first batch when feeding an RNN, the embeddings exist, but after the second batch they don't and I get the following erro

Solution 1:

Seems that the code is trying to create a new Adam Variable in each batch. Possible that the add_training_op is called twice? Also, the snippet of def add_training_op is incomplete since there is no return statement.

Solution 2:

The problem turned out to be the following line of code:

model = RNNLM_Model(config)
    # This instructs gen_model to reuse the same variables as the model above
    scope.reuse_variables()
    gen_model = RNNLM_Model(gen_config)

It turns out that the second model was an issue by using reuse_variables(). By removing this line by issues went away.