tianshou.data.tester

tianshou.data.tester.test_policy_in_env(policy, env, num_timesteps=0, num_episodes=0, discount_factor=0.99, seed=0, episode_cutoff=None)

Tests the policy in the environment and record and prints out the performance. This is useful when the policy is trained with off-policy algorithms and thus the rewards in the data buffer does not reflect the performance of the current policy.

Parameters:

• policy – A tianshou.core.policy. The current policy being optimized.
• env – An environment.
• num_timesteps – An int specifying the number of timesteps to test the policy. It defaults to 0 and either num_timesteps or num_episodes could be set but not both.
• num_episodes – An int specifying the number of episodes to test the policy. It defaults to 0 and either num_timesteps or num_episodes could be set but not both.
• discount_factor – Optional. A float in range \([0, 1]\) defaulting to 0.99. The discount factor to compute discounted returns.
• seed – An non-negative int. The seed to seed the environment as env.seed(seed).
• episode_cutoff – Optional. An int. The maximum number of timesteps in one episode. This is useful when the environment has no terminal states or a single episode could be prohibitively long. If set than all episodes are forced to stop beyond this number to timesteps.