cartpole-impala-tf: env: CartPole-v0 run: IMPALA stop: episode_reward_mean: 150 timesteps_total: 500000 config: # Works for both torch and tf. use_pytorch: false num_gpus: 0