Source code for pfrl.explorers.boltzmann

import numpy as np
import torch
import torch.nn.functional as F

import pfrl

[docs]class Boltzmann(pfrl.explorer.Explorer): """Boltzmann exploration. Args: T (float): Temperature of Boltzmann distribution. """ def __init__(self, T=1.0): self.T = T def select_action(self, t, greedy_action_func, action_value=None): assert action_value is not None assert isinstance(action_value, pfrl.action_value.DiscreteActionValue) n_actions = action_value.q_values.shape[1] with torch.no_grad(): probs = ( F.softmax(action_value.q_values / self.T, dim=-1).cpu().numpy().ravel() ) return np.random.choice(np.arange(n_actions), p=probs) def __repr__(self): return "Boltzmann(T={})".format(self.T)