3f94bc83bf29a7edf887f48f1ece6dc4306d439e,tf_agents/bandits/environments/stationary_stochastic_per_arm_py_environment.py,StationaryStochasticPerArmPyEnvironment,_observe,#StationaryStochasticPerArmPyEnvironment#,123

Before Change

    ], (self._batch_size, self._max_num_actions, -1))
    self._observation = {GLOBAL_KEY: global_obs, PER_ARM_KEY: arm_obs}
    if self._add_mask:
      num_actions = [[self._num_actions_fn()] for _ in range(self._batch_size)]
      num_actions = np.maximum(num_actions, 1)
      mask = np.array(
          np.less(range(self._max_num_actions), num_actions), dtype=np.int32)

After Change

    if self._num_actions_fn:
      num_actions = [self._num_actions_fn() for _ in range(self._batch_size)]
      num_actions = np.maximum(num_actions, 1)
      num_actions = np.minimum(num_actions, self._max_num_actions)
      if self._add_num_actions_feature:
        self._observation.update({NUM_ACTIONS_KEY: num_actions})
      else:
        mask = np.array(
            np.less(range(self._max_num_actions), [[i] for i in num_actions]),
            dtype=np.int32)
        self._observation = (self._observation, mask)
    return self._observation

Instances