From 6aea8c06edb59e174fcb072b3010110f84932d1a Mon Sep 17 00:00:00 2001 From: amandlek Date: Mon, 4 Oct 2021 10:51:40 -0700 Subject: [PATCH 1/2] update roboturk docs with links to robomimic --- docs/algorithms/demonstrations.md | 6 ++++-- docs/algorithms/roboturk.md | 8 +++++--- 2 files changed, 9 insertions(+), 5 deletions(-) diff --git a/docs/algorithms/demonstrations.md b/docs/algorithms/demonstrations.md index cd8f9fea40..ac3e4608df 100644 --- a/docs/algorithms/demonstrations.md +++ b/docs/algorithms/demonstrations.md @@ -45,7 +45,7 @@ We have included an example script that illustrates how demonstrations can be lo We have included some sample demonstrations for each task at `models/assets/demonstrations`. -Our twin project [RoboTurk](http://roboturk.stanford.edu) has also collected pilot datasets of more than a thousand demonstrations for two tasks in our suite via crowdsourcing. You can find detailed information about the RoboTurk datasets [here](roboturk). +Our twin project [RoboTurk](http://roboturk.stanford.edu) has also collected several human demonstration datasets across different tasks and humans, including pilot datasets of more than a thousand demonstrations for two tasks in our suite via crowdsourcing. You can find detailed information about the RoboTurk datasets [here](roboturk). ## Structure of collected demonstrations @@ -81,7 +81,9 @@ The reason for storing mujoco states instead of raw observations is to make it e ## Using Demonstrations for Learning -[Several](https://arxiv.org/abs/1802.09564) [prior](https://arxiv.org/abs/1807.06919) [works](https://arxiv.org/abs/1804.02717) have demonstrated the effectiveness of altering the start state distribution of training episodes for learning RL policies. We provide a generic utility for setting various types of learning curriculums which dictate how to sample from demonstration episodes when doing an environment reset. For more information see the `DemoSamplerWrapper` class. +We have recently released the [robomimic](https://arise-initiative.github.io/robomimic-web/) framework, which makes it easy to train policies using your own [datasets collected with robosuite](https://arise-initiative.github.io/robomimic-web/docs/introduction/datasets.html#robosuite-hdf5-datasets), and other publically released datasets (such as those collected with RoboTurk). The framework also contains many useful examples for how to integrate hdf5 datasets into your own learning pipeline. + +The robosuite repository also has some utilities for using the demonstrations to alter the start state distribution of training episodes for learning RL policies - this have proved effective in [several](https://arxiv.org/abs/1802.09564) [prior](https://arxiv.org/abs/1807.06919) [works](https://arxiv.org/abs/1804.02717). For example, we provide a generic utility for setting various types of learning curriculums which dictate how to sample from demonstration episodes when doing an environment reset. For more information see the `DemoSamplerWrapper` class. ## Warnings We have verified that deterministic action playback works specifically when playing back demonstrations on the *same machine* that the demonstrations were originally collected upon. However, this means that deterministic action playback is NOT guaranteed (in fact, very unlikely) to work across platforms or even across different machines using the same OS. diff --git a/docs/algorithms/roboturk.md b/docs/algorithms/roboturk.md index c60163a0b5..5c5fcc16b2 100644 --- a/docs/algorithms/roboturk.md +++ b/docs/algorithms/roboturk.md @@ -2,11 +2,11 @@ [RoboTurk](https://roboturk.stanford.edu/) is a crowdsourcing platform developed in order to enabled collecting large-scale manipulation datasets. Below, we describe RoboTurk datasets that are compatible with robosuite. -## Updated Datasets compatible with v1.0+ +## Datasets compatible with v1.2+ -We are currently in the process of organizing a standardized dataset for our benchmarking tasks, which will be made available soon and compatible with v1.2.0+. In the meantime, we have provided a [small-scale dataset](https://drive.google.com/drive/folders/1LLkuFnRdqQ6xn1cYzkbJUs_DreaAvN7i?usp=sharing) of expert demonstrations on two of our tasks. +We have collected several human demonstration datasets across several tasks implemented in robosuite as part of the [robomimic](https://arise-initiative.github.io/robomimic-web/) framework. For more information on these datasets, including how to download them and start training policies with them, please see [this link](https://arise-initiative.github.io/robomimic-web/docs/introduction/results.html#downloading-released-datasets). -## Original Datasets compatible with v0.3 +## Datasets compatible with v0.3 We collected a large-scale dataset on the `SawyerPickPlace` and `SawyerNutAssembly` tasks using the [RoboTurk](https://crowdncloud.ai/) platform. Crowdsourced workers collected these task demonstrations remotely. It consists of **1070** successful `SawyerPickPlace` demonstrations and **1147** successful `SawyerNutAssembly` demonstrations. @@ -14,6 +14,8 @@ We are providing the dataset in the hopes that it will be beneficial to research You can download the dataset [here](http://cvgl.stanford.edu/projects/roboturk/RoboTurkPilot.zip). +**Note:** to get started with this data, we highly recommend using the [robomimic](https://arise-initiative.github.io/robomimic-web/) framework - see [this link](https://arise-initiative.github.io/robomimic-web/docs/introduction/datasets.html#roboturk-pilot-datasets) for more information. + After unzipping the dataset, the following subdirectories can be found within the `RoboTurkPilot` directory. - **bins-full** From 1c905de584d5efc517205c75eef610f4e12ab187 Mon Sep 17 00:00:00 2001 From: amandlek Date: Mon, 4 Oct 2021 11:29:33 -0700 Subject: [PATCH 2/2] some minor changes --- docs/algorithms/demonstrations.md | 2 +- docs/algorithms/roboturk.md | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-) diff --git a/docs/algorithms/demonstrations.md b/docs/algorithms/demonstrations.md index ac3e4608df..170f91656d 100644 --- a/docs/algorithms/demonstrations.md +++ b/docs/algorithms/demonstrations.md @@ -45,7 +45,7 @@ We have included an example script that illustrates how demonstrations can be lo We have included some sample demonstrations for each task at `models/assets/demonstrations`. -Our twin project [RoboTurk](http://roboturk.stanford.edu) has also collected several human demonstration datasets across different tasks and humans, including pilot datasets of more than a thousand demonstrations for two tasks in our suite via crowdsourcing. You can find detailed information about the RoboTurk datasets [here](roboturk). +Our sister project [RoboTurk](http://roboturk.stanford.edu) has also collected several human demonstration datasets across different tasks and humans, including pilot datasets of more than a thousand demonstrations for two tasks in our suite via crowdsourcing. You can find detailed information about the RoboTurk datasets [here](roboturk). ## Structure of collected demonstrations diff --git a/docs/algorithms/roboturk.md b/docs/algorithms/roboturk.md index 5c5fcc16b2..723e47d8b3 100644 --- a/docs/algorithms/roboturk.md +++ b/docs/algorithms/roboturk.md @@ -14,7 +14,7 @@ We are providing the dataset in the hopes that it will be beneficial to research You can download the dataset [here](http://cvgl.stanford.edu/projects/roboturk/RoboTurkPilot.zip). -**Note:** to get started with this data, we highly recommend using the [robomimic](https://arise-initiative.github.io/robomimic-web/) framework - see [this link](https://arise-initiative.github.io/robomimic-web/docs/introduction/datasets.html#roboturk-pilot-datasets) for more information. +**Note:** to get started with this data, we highly recommend using the [robomimic](https://arise-initiative.github.io/robomimic-web/) framework - see [this link](https://arise-initiative.github.io/robomimic-web/docs/introduction/datasets.html#roboturk-pilot-datasets) for more information. To use this data, you should be on the [roboturk_v1](https://github.com/ARISE-Initiative/robosuite/tree/roboturk_v1) branch of robosuite, which is `v0.3` with a few minor changes. You can do this by using `git checkout roboturk_v1` after cloning the repository, or just download the source code from [this link](https://github.com/ARISE-Initiative/robosuite/tree/roboturk_v1). After unzipping the dataset, the following subdirectories can be found within the `RoboTurkPilot` directory.