Accelerated Tensorflow on mac


I noticed this announcement - there’s some extremely large increases in performance there: Accelerating TensorFlow Performance on Mac — The TensorFlow Blog

Before diving in, I wondered if anyone can tell me the likelihood of this working with rasa? Many thanks! Duncan

Hi Duncan,

It’s hard to tell at this point! This is all very new. They are using the new macs which we haven’t tried yet. The GitHub - apple/tensorflow_macos: TensorFlow for macOS 11.0+ accelerated using Apple's ML Compute framework. is also very new, and potentially buggy. And the benchmarks in that blog are on strictly conventional nets, different from the architectures used in Rasa.

All that being said though, I think there’s definitely potential. I would say give it a go if you have the time! Would be very interested to hear any results or feedback :slight_smile:

OK, well not a big surprise but it didn’t work - I got a version conflict with Tensorflow, followed by a crash (which I am assuming is related, but not necessarily).

Tensorflow Addons supports using Python ops for all Tensorflow versions above or equal to 2.2.0 and strictly below 2.4.0 (nightly versions are not supported). The versions of TensorFlow you are currently using is 2.4.0-rc0 and is not supported. Some things might work, some things might not.

Still, it’s hopeful that a big performance increase might come our way sometime - the benchmarks are with the new macs (m1 processor) but also using the GPU on a mac pro (intel) machine.

1 Like

Thanks for trying that out @duncsand - and for posting your findings. I also think that eventually we will be able to benefit from this. Once this becomes more mature we will definitely look into it.