[Embench] TinyML MLPerf results

Ray Simar ray.simar at rice.edu
Sat Apr 9 19:14:21 CEST 2022


Hi Dave,

Thanks for this!  I started giving it a cursory look.  I notice especially that some of these have been run on the Arm M4, so my guess is that they might be in the right ball park of fitting into an embedded memory footprint.  I’ll see if I can find out the answer to this more precisely.

Also, I started exploring the possibility of using some of the TinyML work from Pete Warden at Google (now formerly at Google and now at Stanford.)  My students have been creating code for the M4/Arduino environment using the TinyML framework.  So we should be able to compare and contrast the two bodies of work and pick something appropriate for embench.

So, I am hopeful that we can either use MLPerfTiny or something from the TinyML framework.  Joe and I can see if we can get some students interested in taking a look this summer.  Stay tuned.

All the best,
Ray

> On Apr 6, 2022, at 12:47 PM, David PATTERSON <pattrsn at cs.berkeley.edu> wrote:
> 
> https://mlcommons.org/en/news/mlperf-inference-1q2022/ <https://urldefense.com/v3/__https://mlcommons.org/en/news/mlperf-inference-1q2022/__;!!BuQPrrmRaQ!lx193PKnPinH2hDyycqFLcmWJNazoacjNdVBryl3Eh7f2z37M-7lWo6DxcPQTTOSNVtG-Xn-oxTaQofLIakJtVbZVvnYaDme$>
> 
> The MLPerf Tiny benchmark suite is intended for the lowest power devices and smallest form factors, such as deeply embedded, intelligent sensing, and internet-of-things applications. The second round of MLPerf Tiny results showed tremendous growth in collaboration with submissions from Alibaba, Andes, hls4ml-FINN team, Plumerai, Renesas, Silicon Labs, STMicroelectronics, and Syntiant. Collectively, these organizations submitted 19 different systems with 3X more results than the first round and over half the results incorporating energy measurements, an impressive achievement for the first benchmarking round with energy measurement.
> To view the results and find additional information about the benchmarks please visit https://mlcommons.org/en/inference-datacenter-20/ <https://urldefense.com/v3/__https://mlcommons.org/en/inference-datacenter-20/__;!!BuQPrrmRaQ!lx193PKnPinH2hDyycqFLcmWJNazoacjNdVBryl3Eh7f2z37M-7lWo6DxcPQTTOSNVtG-Xn-oxTaQofLIakJtVbZVr3BM2T7$>,
>  https://www.mlcommons.org/en/inference-tiny-07/ <https://urldefense.com/v3/__https://www.mlcommons.org/en/inference-tiny-07/__;!!BuQPrrmRaQ!lx193PKnPinH2hDyycqFLcmWJNazoacjNdVBryl3Eh7f2z37M-7lWo6DxcPQTTOSNVtG-Xn-oxTaQofLIakJtVbZVmihkpZU$>-- 
> Embench mailing list
> Embench at lists.librecores.org
> https://urldefense.com/v3/__https://lists.librecores.org/listinfo/embench__;!!BuQPrrmRaQ!lx193PKnPinH2hDyycqFLcmWJNazoacjNdVBryl3Eh7f2z37M-7lWo6DxcPQTTOSNVtG-Xn-oxTaQofLIakJtVbZVqqJjy15$ 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.librecores.org/pipermail/embench/attachments/20220409/a06aac35/attachment.htm>


More information about the Embench mailing list