tensorflow c++ model from keras model doesnt have "init" operation

[OfekShochat]

hi, I dont know if this is good for this forum, but I ran into a problem when programming the network in my chess engine.
I use the cppflow c_api wrapper and I converted my model to a pb file with https://medium.com/@sebastingarcaacosta ... 740846d9eb and it works. when I try to initialize it says that there is no init operation and indeed when I get the operations it doesn't specify something close to "init". and when I do `tf.compat.v1.variables_initializer([])` it returns None.
does anyone know how to get an init operation in?

[Karlo Bala]

hi, I dont know if this is good for this forum, but I ran into a problem when programming the network in my chess engine.
I use the cppflow c_api wrapper and I converted my model to a pb file with https://medium.com/@sebastingarcaacosta ... 740846d9eb and it works. when I try to initialize it says that there is no init operation and indeed when I get the operations it doesn't specify something close to "init". and when I do `tf.compat.v1.variables_initializer([])` it returns None.
does anyone know how to get an init operation in?

I'm using cppflow2 for car recognition, and basically, I need only two operations, load net, and forward pass. It is enough to save net in a folder from Keras (no convert or whatsoever), and later just reload the model from cppflow2. Now, I'm not sure what are you trying to achieve.