f-IRL
Meta-SAC
f-IRL | Meta-SAC | |
---|---|---|
2 | 1 | |
35 | 28 | |
- | - | |
1.8 | 0.0 | |
10 months ago | almost 3 years ago | |
Python | Python | |
MIT License | MIT License |
Stars - the number of stars that a project has on GitHub. Growth - month over month growth in stars.
Activity is a relative number indicating how actively a project is being developed. Recent commits have higher weight than older ones.
For example, an activity of 9.0 indicates that a project is amongst the top 10% of the most actively developed projects that we are tracking.
f-IRL
-
Can you use the reward learned in generative adversarial imitation learning in order to train from scratch?
Code for https://arxiv.org/abs/2011.04709 found: https://github.com/twni2016/f-IRL
-
How to pretrain a model on expert data?
Try using an imitation learning algorithm. Two popular options are MaxEnt IRL and GAIL. This repository has GAIL implementation and this repository has MaxEnt IRL and GAIL implementation. There are other implementations too that you can check out.
Meta-SAC
-
Do policy gradient methods also require some mechanism for exploration?
A simple approach that can help is a linear entropy schedule: Start at a high value to explore early, decay over time to learn a more optimal policy. Some variants of SAC autotune the entropy over time. A more advanced approach is AGAC, which does something like a GAN to encourage the PPO/A2C policy to explore by forcing it to be less predictable. There are many approaches, these are just a sample
What are some alternatives?
falken - Falken provides developers with a service that allows them to train AI that can play their games
autonomous-learning-library - A PyTorch library for building deep reinforcement learning agents.
ElegantRL - Massively Parallel Deep Reinforcement Learning. 🔥
tianshou - An elegant PyTorch deep reinforcement learning library.
eirli - An Empirical Investigation of Representation Learning for Imitation (EIRLI), NeurIPS'21
pytorch-a2c-ppo-acktr-gail - PyTorch implementation of Advantage Actor Critic (A2C), Proximal Policy Optimization (PPO), Scalable trust-region method for deep reinforcement learning using Kronecker-factored approximation (ACKTR) and Generative Adversarial Imitation Learning (GAIL).
imitation - Clean PyTorch implementations of imitation and reward learning algorithms