Decision Transformer: A brand new Offline RL Pattern

Overview

DecisionTransformer_StepbyStep

Intro

Decision Transformer: A brand new Offline RL Pattern.

这是关于NeurIPS 2021 热门论文Decision Transformer的复现。

👍 原文地址: Decision Transformer: Reinforcement Learning via Sequence Modeling

👍 官方的Git仓库: decision-transformer(official)

Decision Transformer

Decision Transformer属于Offline RL,所谓Offline RL,即从次优数据中学习策略来分配Agent,即从固定、有限的经验中产生最大有效的行为。

👀️ Motivation

DT将RL看成一个序列建模问题(Sequence Modeling Problem ),不用传统RL方法,而使用网络直接输出动作进行决策。传统RL方法存在一些问题,比如估计未来Return过程中Bootstrapping过程会导致Overestimate; 马尔可夫假设;

DT借助了Transformer的强大表征能力和时序建模能力。

  • Decision Transformer的表现达到甚至超过了目前最好的基于dynamic programming的主流方法;
  • 在一些需要long-term credit assignment的task【例如sparse reward或者delayed reward等】,Decision Transformer的表现远超过了最好的主流方法.

🚀️ DT的核心思想

image.png

Decision Transformer的核心思想; States、Actions、Returns被Fed into Modality-Specific的线性Embedding;并添加了带有时间步信息的positional episodic timestep; 这些Tokens被输入一个GPT架构,使用a causal self-attention mask来预测actions。

🎉️ DT的优势

  1. 无需Markov假设;
  2. 没有使用一个可学习的Value Function作为Training Target;
  3. 利用Transformer的特性,绕过长期信用分配进行“自举bootstrapping”的需要,避免了时序差分学习的“短视”行为;
  4. 可以通过self-attention直接执行信度分配。这与缓慢传播奖励并容易产生干扰信号的 Bellman Backup 相反,可以使 Transformer 在奖励稀少或分散注意力的情况下仍然有效地工作.

Dependencies

1. D4RL ( Dataset for Deep Data-Driven Reinforcement Learning )

2. MUJOCO 210

# 安装之前先安装absl-py和matplotlib 
pip install absl-py 
pip install matplotlib 

"""
git clone https://github.com/rail-berkeley/d4rl.git
cd d4rl
pip install -e . # 这种方法不好使 !! 
"""

#首先在https://github.com/deepmind/dm_control这个库git clone
# cd
pip install -r requirement.txt 
# 然后 
pip install matplotlib 
# 然后 https://github.com/takuseno/d3rlpy 
pip install d3rlpy 
# 然后安装mujoco 210  
# 直接安装,然后添加环境变量 
# 装完之后进d4rl文件夹下
python setup.py install 
# 成功安装 d4rl 1.1 

3. GPT-2


pip install transformers

Experiments

Group1: Decision Transformer — Hopper-v3-Medium-Dataset

参数Config

class Config:
    env = "hopper"
    dataset = "medium"
    mode = "normal" # "delayed" : all rewards moved to end of trajectory
    device = 'cuda'
    log_dir = 'TB_log/'
    record_algo = 'DT_Hopper_v1'
    test_cycles = datetime.datetime.now().strftime('%Y%m%d_%H%M%S')

    # 模型
    model_type = "DT"
    activation_function = 'relu'

    # Scalar
    max_length = 20 # max_len # K
    pct_traj = 1.
    batch_size = 64
    embed_dim = 128
    n_layer = 3
    n_head = 1
    dropout = 0.1
    lr = 1e-4
    wd = 1e-4
    warmup_steps = 1000
    num_eval_episodes = 100
    max_iters = 50
    num_steps_per_iter = 1000

    # Bool
    log_to_tb = True

效果

image.png

Owner
Irving
Irving
Implementation of "Deep Implicit Templates for 3D Shape Representation"

Deep Implicit Templates for 3D Shape Representation Zerong Zheng, Tao Yu, Qionghai Dai, Yebin Liu. arXiv 2020. This repository is an implementation fo

Zerong Zheng 144 Dec 07, 2022
ATAC: Adversarially Trained Actor Critic

ATAC: Adversarially Trained Actor Critic Adversarially Trained Actor Critic for Offline Reinforcement Learning by Ching-An Cheng*, Tengyang Xie*, Nan

Microsoft 41 Dec 08, 2022
A PyTorch re-implementation of the paper 'Exploring Simple Siamese Representation Learning'. Reproduced the 67.8% Top1 Acc on ImageNet.

Exploring simple siamese representation learning This is a PyTorch re-implementation of the SimSiam paper on ImageNet dataset. The results match that

Taojiannan Yang 72 Nov 09, 2022
Official Implementation of "LUNAR: Unifying Local Outlier Detection Methods via Graph Neural Networks"

LUNAR Official Implementation of "LUNAR: Unifying Local Outlier Detection Methods via Graph Neural Networks" Adam Goodge, Bryan Hooi, Ng See Kiong and

Adam Goodge 25 Dec 28, 2022
RepVGG: Making VGG-style ConvNets Great Again

RepVGG: Making VGG-style ConvNets Great Again (PyTorch) This is a super simple ConvNet architecture that achieves over 80% top-1 accuracy on ImageNet

2.8k Jan 04, 2023
1st-in-MICCAI2020-CPM - Combined Radiology and Pathology Classification

Combined Radiology and Pathology Classification MICCAI 2020 Combined Radiology a

22 Dec 08, 2022
Pytorch implementation of VAEs for heterogeneous likelihoods.

Heterogeneous VAEs Beware: This repository is under construction 🛠️ Pytorch implementation of different VAE models to model heterogeneous data. Here,

Adrián Javaloy 35 Nov 29, 2022
This is a Python Module For Encryption, Hashing And Other stuff

EnroCrypt This is a Python Module For Encryption, Hashing And Other Basic Stuff You Need, With Secure Encryption And Strong Salted Hashing You Can Do

5 Sep 15, 2022
Pytorch implementation of the paper SPICE: Semantic Pseudo-labeling for Image Clustering

SPICE: Semantic Pseudo-labeling for Image Clustering By Chuang Niu and Ge Wang This is a Pytorch implementation of the paper. (In updating) SOTA on 5

Chuang Niu 154 Dec 15, 2022
Source code for "FastBERT: a Self-distilling BERT with Adaptive Inference Time".

FastBERT Source code for "FastBERT: a Self-distilling BERT with Adaptive Inference Time". Good News 2021/10/29 - Code: Code of FastPLM is released on

Weijie Liu 584 Jan 02, 2023
Scaling Vision with Sparse Mixture of Experts

Scaling Vision with Sparse Mixture of Experts This repository contains the code for training and fine-tuning Sparse MoE models for vision (V-MoE) on I

Google Research 290 Dec 25, 2022
Official implementation of "OpenPifPaf: Composite Fields for Semantic Keypoint Detection and Spatio-Temporal Association" in PyTorch.

openpifpaf Continuously tested on Linux, MacOS and Windows: New 2021 paper: OpenPifPaf: Composite Fields for Semantic Keypoint Detection and Spatio-Te

VITA lab at EPFL 50 Dec 29, 2022
SMORE: Knowledge Graph Completion and Multi-hop Reasoning in Massive Knowledge Graphs

SMORE: Knowledge Graph Completion and Multi-hop Reasoning in Massive Knowledge Graphs SMORE is a a versatile framework that scales multi-hop query emb

Google Research 135 Dec 27, 2022
PyBullet CartPole and Quadrotor environments—with CasADi symbolic a priori dynamics—for learning-based control and reinforcement learning

safe-control-gym Physics-based CartPole and Quadrotor Gym environments (using PyBullet) with symbolic a priori dynamics (using CasADi) for learning-ba

Dynamic Systems Lab 300 Dec 28, 2022
Prototypical Cross-Attention Networks for Multiple Object Tracking and Segmentation, NeurIPS 2021 Spotlight

PCAN for Multiple Object Tracking and Segmentation This is the offical implementation of paper PCAN for MOTS. We also present a trailer that consists

ETH VIS Group 328 Dec 29, 2022
This is the official pytorch implementation for the paper: Instance Similarity Learning for Unsupervised Feature Representation.

ISL This is the official pytorch implementation for the paper: Instance Similarity Learning for Unsupervised Feature Representation, which is accepted

19 May 04, 2022
Official Pytorch implementation of the paper "MotionCLIP: Exposing Human Motion Generation to CLIP Space"

MotionCLIP Official Pytorch implementation of the paper "MotionCLIP: Exposing Human Motion Generation to CLIP Space". Please visit our webpage for mor

Guy Tevet 173 Dec 26, 2022
Code for ICCV 2021 paper Graph-to-3D: End-to-End Generation and Manipulation of 3D Scenes using Scene Graphs

Graph-to-3D This is the official implementation of the paper Graph-to-3d: End-to-End Generation and Manipulation of 3D Scenes Using Scene Graphs | arx

Helisa Dhamo 33 Jan 06, 2023
Wanli Li and Tieyun Qian: Exploit a Multi-head Reference Graph for Semi-supervised Relation Extraction, IJCNN 2021

MRefG Wanli Li and Tieyun Qian: "Exploit a Multi-head Reference Graph for Semi-supervised Relation Extraction", IJCNN 2021 1. Requirements To reproduc

万理 5 Jul 26, 2022
The code for paper Efficiently Solve the Max-cut Problem via a Quantum Qubit Rotation Algorithm

Quantum Qubit Rotation Algorithm Single qubit rotation gates $$ U(\Theta)=\bigotimes_{i=1}^n R_x (\phi_i) $$ QQRA for the max-cut problem This code wa

SheffieldWang 0 Oct 18, 2021