PortfolioEnv

class PortfolioEnv(gym.Env):
  def _step(self, action):
    """
    Step the env.
    Actions should be portfolio [w0...]
    - Where wn is a portfolio weight between 0 and 1. The first (w0) is cash_bias
    - cn is the portfolio conversion weights see PortioSim._step for description
    """
    logger.debug('action: %s', action)

    weights = np.clip(action, 0.0, 1.0)
    weights /= weights.sum() + eps

    # Sanity checks
    assert self.action_space.contains(
        action), 'action should be within %r but is %r' % (self.action_space, action)
    np.testing.assert_almost_equal(
        np.sum(weights), 1.0, 3, err_msg='weights should sum to 1. action="%s"' % weights)

    history, y1, done1 = self.src._step()

    reward, info, done2 = self.sim._step(weights, y1)

    # calculate return for buy and hold a bit of each asset
    info['market_value'] = np.cumprod(
        [inf["market_return"] for inf in self.infos + [info]])[-1]
    # add dates
    info['date'] = self.src.times[self.src.step].timestamp()
    info['steps'] = self.src.step

    self.infos.append(info)

    # reshape history according to output mode
    if self.output_mode == 'EIIE':
        pass
    elif self.output_mode == 'atari':
        padding = history.shape[1] - history.shape[0]
        history = np.pad(history, [[0, padding], [
            0, 0], [0, 0]], mode='constant')
    elif self.output_mode == 'mlp':
        history = history.flatten()

    return {'history': history, 'weights': weights}, reward, done1 or done2, info