Home

Welcome to the DRL_Implementation wiki!

This tutorial will walk you through the usages, building blocks and programing logic of this Pytorch-based Deep Reinforcement Learning implementation package.

1. Using off-the-shell agent.
2. What's in the package?
   1. Overview of the building blocks.
   2. Base agent class.
   3. Utility classes.

After the tutorial, you can proceed to create your own agent using common components of modern DRL algorithms available from this package. Feel free to raise issues. Here are some tips that may help:

• It is recommended to start by modifying an existing agent.
• For compatibility, make sure your networks and statistics are saved in dictionaries.

Here's a starting example:

import time
import numpy as np
import torch as T
import torch.nn.functional as F
from torch.optim.adam import Adam
from drl_implementation.agent.agent_base import Agent
from drl_implementation.agent.utils.networks_mlp import Actor, Critic
from drl_implementation.agent.utils.exploration_strategy import GaussianNoise

class MyDDPG(Agent):
    def __init__(self, ...):
        ...