Infer.NET Team Blog - All Comments

re: Parallel inference in Infer.NET 2.3

shakey — Wed, 25 Nov 2009 06:19:37 GMT

the Microsoft Parallel Extensions CTP link is broken, I download the System.Threading.dll file but fail to get infer.net to recognise the parallel type in it. any advice?

re: Parallel inference in Infer.NET 2.3

NeillC — Thu, 29 Oct 2009 17:10:05 GMT

I have emailed the code (basically the GMM fitting example) and a test data file if you have a chance to check it out :-) - I hope it's not too troublesome.

Cheers, Neill

re: Parallel inference in Infer.NET 2.3

jwinn — Thu, 29 Oct 2009 10:02:06 GMT

NeillC:

I'm sorry this isn't working for you - it looks like the code generated internally by Infer.NET is defeating the C# compiler. Can you post your model code on the Infer.NET forums (or email it to infersup@microsoft.com), so that I can reproduce the problem and try to work out what is going wrong?

Thanks, John

re: Parallel inference in Infer.NET 2.3

NeillC — Tue, 27 Oct 2009 18:21:44 GMT

Hi John,

I've been using the VMP algorithm to fit a GMM to a 7D data set and it is woking well :-) but quite slow. I'm very keen to use the parallel extensions, which I hope will speed things up, but I get the following error message:

Compiling model...done.

Compilation time was 2605ms.

Compilation failed with 1 error(s)

error CS1647: An expression is too long or complex to compile near 'MicrosoftResearch.Infer.Models.User.Model_VMP.Reset()'

Do you know if I'm doing something stupid or rather being too ambitious with the model (~8000 data points in 7 dimensions)? It works correctly if I just remove the ie.Compiler.UseParallelForLoops = true; line.

Thanks,

Neill

re: Parallel inference in Infer.NET 2.3

jwinn — Fri, 14 Aug 2009 10:59:08 GMT

I've not done extensive tests on this, but informally it seems the memory costs are small - presumably dominated by the memory overhead of introducing additional threads in the thread pool e.g. the stack for each thread.

re: Parallel inference in Infer.NET 2.3

laura — Fri, 14 Aug 2009 10:32:42 GMT

This is excellent!

How much does parallel inference cost in terms of memory?

Cheers,

Laura

re: Infer.NET released!!

jwinn — Thu, 18 Dec 2008 16:55:42 GMT

Steven,

Thanks for the positive comments!

We do not have built-in support for running inference on the GPU. However, it would be possible to use the Infer.NET extension mechanism to do this, as described here:

research.microsoft.com/.../How%20to%20add%20a%20new%20factor%20and%20message%20operators.aspx

The idea is that you can insert customised message operators (the computations that act on the messages during message passing). These can be used to support new types of factor, or different implementations of existing messages e.g. using the GPU or other distributed computing mechanisms. In order to get the maximum speedup, you would want to create factors which operate on whole arrays of messages rather than individual messages - otherwise any gains would be lost in the overhead of invoking the GPU.

It would also be possible to eliminate most transfers to/from the GPU by using proxy message classes which would hold a pointer to an array of messages stored in the GPU. Message operations on such messages would then operate entirely within the GPU with no data transfer required.

I would be happy to discuss this in more detail. If you would like to do so, then please start a forum thread on this topic.

Best wishes,

John

re: Infer.NET released!!

StevenLongmire — Fri, 12 Dec 2008 13:35:47 GMT

Hello Infer.NET team!

Awesome developments with your efforts!

Is there a way to "offload" into CUDA and Tesla the HMM modeling for mass data set inference?

I have been looking into this the last couple of years for text ontologies and recognition of context meanings in language for threat based scenairos. Also, for use in AI on IT environment threat research.

Any info would be apperciated.

steven@stevenlongmire.com