More information about the Underscore mailing list

[_] AWS ML

Matt Hamilton matt at quernus.co.uk
Wed Feb 20 11:03:11 GMT 2019

Yes, you can. My point is that GPUs don't often speed up RL work. I was curious since reinforcement learning was mentioned specifically is AWS had any specific tools for helping with RL.

-Matt

> On 20 Feb 2019, at 10:01, David Hodgkinson <davehodg at gmail.com> wrote:
> 
> You can get GPUs on AWS, no? CUDA FTW!
> 
>> On 20 Feb 2019, at 09:41, Matt Hamilton <matt at quernus.co.uk> wrote:
>> 
>> Very interesting... do AWS have anything specific to help reinforcement learning? I've generally not had much luck running that on cloud servers as never seem to get much speedup. I think the overhead of loading on and off GPUs outweighs any performance gains you get.  I have custom compiled tensorflow with AVX and MMX2 on my iMac (3.2Ghz Core i5) and it seems to perform better than any behemoth I've managed to run on AWS/GCE.
>> 
>> -Matt
>> 
>>> On 20 Feb 2019, at 09:03, Adrian H <rozangela at hotmail.com> wrote:
>>> 
>>> I am in Bristol, company overseas. We are looking to become an AWS ML partner, but do not yet have enough case studies using this technology stack. So, if you have a need for reinforcement learning, AI, ML or a data scientist and are happy we build the solution within AWS. I am sure we can cut a really good deal. 
>>> 
>>> Let me know
>>> 
>>> Adrian
>>> -- 
>>> underscore_ list info/archive -> http://www.under-score.org.uk/mailman/listinfo/underscore
>> 
>> 
>>>> Matt Hamilton
>> Quernus
>> matt at quernus.co.uk
>> +44 117 325 3025
>> 
>> Quernus Ltd is a company registered in England and Wales. Registered number: 09076246
>> 
>> -- 
>> underscore_ list info/archive -> http://www.under-score.org.uk/mailman/listinfo/underscore
> -- 
> underscore_ list info/archive -> http://www.under-score.org.uk/mailman/listinfo/underscore


— 
Matt Hamilton
Quernus
matt at quernus.co.uk
+44 117 325 3025

Quernus Ltd is a company registered in England and Wales. Registered number: 09076246