loaded keras model doesn't have predict method

Question

In the DQN algorithm, the agent is trained successfully and the model is saved as a .h5 file. The agent class has a method called load as follows:

def load(self):
    new_model = load_model('dqn_model.h5')
    if self.epsilon == 0.0:
       self.update_network_parameters()
    return new_model

def update_network_parameters(self):
    self.model.set_weights(self.new_model.get_weights())

I try to run the evaluation step by creating the new_model as follows:

new_model = agent.load()

but the new_model doesn't have any method like predict.

So you can't run `new_model.predict()`. Apologies as I am unfamiliar with keras, but is load_model something from keras? — user1558604, Mar 17 '21 at 02:37
How were you trying to run predict? You wouldn't run it on the `load` object, but instead the model that you loaded (`new_model`) — user1558604, Mar 17 '21 at 02:47
That seems unlikely to me given that the documentation says that `load_model` returns a `Model` instance and that `Model` instances have a predict method. Can you print the results of `dir(new_model)` and include that in the question? — user1558604, Mar 17 '21 at 02:53

score 0 · Answer 1 · answered Mar 17 '21 at 03:00

Ensure you have Model imported within the class your calling agent.load from.

from keras.models import Model

Also, for me personally, model.predict doesn't show up within Intellisense so make sure that code actually doesn't compile when you run it. If so tell us what the exception error is.

loaded keras model doesn't have predict method

1 Answers1