CODEWITHSUNDEEP
pythondeep-learningpytorchreinforcement-learning
ke qi
ke qi

Reputation: 44

How to print the adjusting learning rate in Pytorch?

While I use torch.optim.Adam and exponential decay_lr in my PPO algorithm:

        self.optimizer = torch.optim.Adam([
            {'params': self.policy.actor.parameters(), 'lr': lr_actor},
            {'params': self.policy.critic.parameters(), 'lr': lr_critic}
        ])
        self.scheduler = torch.optim.lr_scheduler.ExponentialLR(self.optimizer, self.GAMMA)

The initial lr=0.1, and GAMMA=0.9.

Then I print the lr in my epoch dynamiclly with:

            if time_step % update_timestep == 0:
                ppo_agent.update()
                print(f'__________start update_______________')
                print(ppo_agent.optimizer.state_dict()['param_groups'][0]['lr'])

But, something wrong with this, and the bug is:

  File "D:\Anaconda\lib\site-packages\torch\distributions\beta.py", line 36, in __init__
    self._dirichlet = Dirichlet(concentration1_concentration0, validate_args=validate_args)
  File "D:\Anaconda\lib\site-packages\torch\distributions\dirichlet.py", line 52, in __init__
    super(Dirichlet, self).__init__(batch_shape, event_shape, validate_args=validate_args)
  File "D:\Anaconda\lib\site-packages\torch\distributions\distribution.py", line 53, in __init__
    raise ValueError("The parameter {} has invalid values".format(param))
ValueError: The parameter concentration has invalid values

Then, if I delete the "print()" sentence, it does work well! So, it's bothering me very much.

Upvotes: 0

Views: 1627

Answers (1)

Theodor Peifer
Theodor Peifer

Reputation: 3496

You can get the learning rate like this:

self.optimizer.param_groups[0]["lr"]

Upvotes: 1

Related Questions