Marco Nardelli

Hi bmac, thank you very much, I'm gonna try this today and let you know if it works. appreciated :)

I'm using a DQN, running eagerly this function seems to work by its own: ``` tf.config.experimental_run_functions_eagerly(True) # constraint for action @tf.function def observation_and_action_constrain_splitter(observation): mask = np.zeros(tf.shape(observation)[1]).astype(dtype=np.int32) line_number = 0 for...

I'm getting the error on the epsilon_greedy_policy file, line 102 `action = tf.compat.v1.where(cond, greedy_action.action, random_action.action)` same line...

I made the same modification to the random_tf_policy and it works fine now :)

Hi Sergio, thank you for your reply, but if I generate it on the PyEnv, how can I pass it to the network? is there any example for this?

Hi bmac, I think the solution for this is to create a custom Network that can receive the dict as a parameter, I'm working on this too...I'll give you an...

Hi bmac, I will try to describe all steps you have to make to get things working. First of all you have to change your environment: Observation spec must be...