number of rows (= actions) returned by get_action in SbAgent

[rahulsavani]

The old method

    # return self.model.predict(state, deterministic=True)[0].reshape(self.num_trajectories, self.num_actions)

used self.num_trajectories as the number of different actions to return, but this was failing in some cases due to a discrepancy between self.num_tracjectories and the actual state given to model predict

Leandro and Rahul changed this to

    return self.model.predict(state, deterministic=True)[0].reshape(state.shape[0], self.num_actions)

Are we happy with this?