21fa4e7e38567e083f98c559f8f2ae0c5573ba0f,ch15/lib/model.py,AgentA2C,__call__,#AgentA2C#,56

Before Change

        mu_v, var_v = self.net(states_v)
        mu = mu_v.data.cpu().numpy()
        sigma = torch.sqrt(var_v).data.cpu().numpy()
        actions = np.random.normal(mu, sigma)
        actions = np.clip(actions, -1, 1)
        return actions, agent_states

After Change

        mu_v = self.net(states_v)
        mu = mu_v.data.cpu().numpy()
        logstd = self.net.logstd.data.cpu().numpy()
        actions = mu + np.exp(logstd) * np.random.normal(size=logstd.shape)
        actions = np.clip(actions, -1, 1)
        return actions, agent_states

Instances