Implementing an Open Source Agent57

Having a YouTube channel has brought me a number of fantastic opportunities over the years. One of my favorites was working on a project to implement an open source agent 57. In this video I'll talk a little about how A57 works and how I worked with a team of Ukranian developers to implement an open source version.

If you want to contribute to the project, check it out here:

You can find a written version of this article with some more details here:

Want to support my work and learn how to turn deep reinforcement learning papers into code?

Get instant access to all my courses, including the new Prioritized Experience Replay course, with my subscription service. $29 a month gives you instant access to 42 hours of instructional content plus access to future updates, added monthly.

Or, pickup my Udemy courses here:

Deep Q Learning:

Actor Critic Methods:

Curiosity Driven Deep Reinforcement Learning

Natural Language Processing from First Principles:

Just getting started in deep reinforcement learning? Check out my intro level course through Manning Publications.

Reinforcement Learning Fundamentals

Here are some books / courses I recommend (affiliate links):

Come hang out on Discord here:

[ChrisPoulin]

Great job, as always Phil! Thanks for co-authoring/co-leading this effort. Absolutely couldn't have done it without you. Thanks, Chris

[hidrogenodeuterio]

Muchisimas gracias por este contenido de gran valor ❤

[manelmimi]

Thanks great job !! Please Can you make a video to explain QMIX algorithm ?

[Biedropegaz]

Phil, text on the slides are hard to read, because window with You is too big :-).

[alignedfibers]

Phil, only gpus I can really get are a couple of AMD RX580 8GB and one of them is for vm passthrough only, so really only direct access to one at a time (there are ways of course) and only have an 4core xeon 8threads 3.5GHZ which has an igpu that is not used at all, 32GB ECC 1600 MHZ (overclockable if needed), nvme shares 16 lanes with gpu, so only 8x lanes gen3 to gpu, I can connect a workstation class gpu to the 4x I have left for video only, so can dedicate that one 580 to compute, but prefer to keep that one slot for network. Will I be able to run any of the lessons you have demonstrate on youtube? I think I have to turn off igpu turbo from motherboard dip switch first, but Maybe I can get some compute loads to run against the igpu also? (I could use both cards from the host but would love just still leave that one passed through. So my system is 20 lanes max on cpu but has additional 8 lanes (electrically four lanes I think) connected to the chipset and shares with usb excetera and that is what the vm uses. Everything will run bare metal, ubuntu,

[FarrukhTickets]

Can you share some thoughts on Chat GPT. I know it is a buzz word these days and everyone is talking about it, but will be great to hear your thoughts.

[karolstylok]

What gpu and motherboard should I buy if I want to train lstm or rnn? My budget is about 2k - 3k usd. I use keras and tensorflow. Where can I find information or what are key informations should I focus on?

[marcopleines]

Is it possible to configure the training to just run plain R2D2?

[eduardorosentreter]

Hello, I'm having difficulties running docker-compose, it's my first time on Linux, I decided to change to a real operating system but I don't know if I'm doing things right or not, when I do docker-compose it's telling me that does not find .env, that file is not in the directory.