c87a29c49f488ce290c5942d964bde73c3ff2c91,samples/dqn_expreplay.py,,,#,15

Before Change

    test_s = Variable(torch.from_numpy(np.array([env.reset()], dtype=np.float32)))
    print(model(test_s))
    print(action_selector(model(test_s)))
    print(loss_fn(model(test_s), Variable(torch.Tensor([[1.0, 0.0, 2.0]]))))

    // TODO: implement env pooling
    exp_replay = ExperienceReplayBuffer(env, params, buffer_size=100)

After Change

    // populate buffer
    exp_replay.populate(50)
    // sample batch from buffer
    batch = exp_replay.sample(50)
    // convert experience batch into training samples
    

    pass

Instances