tf_agents.bandits.policies.policy_utilities.InfoFields

Strings which can be used in the policy info fields.

BANDIT_POLICY_TYPE 'bandit_policy_type'
CHOSEN_ARM_FEATURES 'chosen_arm_features'
LOG_PROBABILITY 'log_probability'
PREDICTED_REWARDS_MEAN 'predicted_rewards_mean'
PREDICTED_REWARDS_OPTIMISTIC 'predicted_rewards_optimistic'
PREDICTED_REWARDS_SAMPLED 'predicted_rewards_sampled'