mirror of
https://github.com/wassname/kair_algorithms_draft.git
synced 2026-08-30 11:25:30 +08:00
Whi Kwon
5962ff5277
* Add td3 * Fix flake8 * Fix action clamping * Increase episode max step, add detach to actor_loss * Fix actor update freq bug * Add per (#8) * Add per and modify etc * Replace pre-commit-config.yaml and add pre-commit hook in .git * Modify .gitignore * Modify .gitignore * Modify buffer and code * Modify replay buffer and per * Modify .gitignore * Add random initial action in ddpg (#13) * Add random initial actions in ddpg * Add reacher-v2 example of ddpg * Add soft actor critic (#12) * Add soft actor critic * Delete unnecessary examples * Add td3 * Fix flake8 * Fix action clamping * Fix flake8 * Increase episode max step, add detach to actor_loss * Fix actor update freq bug * Fix code to reflect PR * Resolve conflict
67 lines
1.8 KiB
Python
67 lines
1.8 KiB
Python
# -*- coding: utf-8 -*-
|
|
"""Noise classes for algorithms."""
|
|
|
|
import copy
|
|
import random
|
|
|
|
import numpy as np
|
|
|
|
|
|
class GaussianNoise:
|
|
"""Gaussian Noise.
|
|
|
|
Taken from https://github.com/vitchyr/rlkit
|
|
"""
|
|
|
|
def __init__(
|
|
self,
|
|
action_dim: int,
|
|
min_sigma: float = 1.0,
|
|
max_sigma: float = 1.0,
|
|
decay_period: int = 1000000,
|
|
):
|
|
"""Initialization."""
|
|
self.action_dim = action_dim
|
|
self.min_sigma = min_sigma
|
|
self.max_sigma = max_sigma
|
|
self.decay_period = decay_period
|
|
|
|
def sample(self, t: int = 0) -> float:
|
|
"""Get an action with gaussian noise."""
|
|
sigma = self.max_sigma - (self.max_sigma - self.min_sigma) * min(
|
|
1.0, t / self.decay_period
|
|
)
|
|
return np.random.normal(0, sigma, size=self.action_dim)
|
|
|
|
|
|
class OUNoise:
|
|
"""Ornstein-Uhlenbeck process.
|
|
|
|
Taken from Udacity deep-reinforcement-learning github repository:
|
|
https://github.com/udacity/deep-reinforcement-learning/blob/master/
|
|
ddpg-pendulum/ddpg_agent.py
|
|
"""
|
|
|
|
def __init__(
|
|
self, size: int, mu: float = 0.0, theta: float = 0.15, sigma: float = 0.2
|
|
):
|
|
"""Initialize parameters and noise process."""
|
|
self.state = np.float64(0.0)
|
|
self.mu = mu * np.ones(size)
|
|
self.theta = theta
|
|
self.sigma = sigma
|
|
self.reset()
|
|
|
|
def reset(self):
|
|
"""Reset the internal state (= noise) to mean (mu)."""
|
|
self.state = copy.copy(self.mu)
|
|
|
|
def sample(self) -> float:
|
|
"""Update internal state and return it as a noise sample."""
|
|
x = self.state
|
|
dx = self.theta * (self.mu - x) + self.sigma * np.array(
|
|
[random.random() for _ in range(len(x))]
|
|
)
|
|
self.state = x + dx
|
|
return self.state
|