When i try to restore a model with use_attention=True i get:

ValueError: Must pass in RNN state batches for placeholders [<tf.Tensor ‘default_policy/Placeholder:0’ shape=(?, ?, 64) dtype=float32>], got []

when I try:

state = [np.zeros([64], dtype=np.float32)]

while not done:

action = agent.compute_action(obs, state)

obs, reward, done, info = env.step(action)

Cannot feed value of shape (1, 64) for Tensor ‘default_policy/Placeholder:0’, which has shape ‘(?, ?, 64)’

how can I create an empty state with shape (?,?,64)?

np.zeros((0,0,64)) does not work.