Missing support for M1 Mac

[utkarshagarwal]

Installation fails on M1 mac for tfx-bsl, tft and tfdv.

[pindinagesh]

@utkarshagarwal

Sorry for late response.

Have you tried with this installation guide? and M1 Macs are not supported for tfx-bsl as it mentioned here, correct me if I am missing something. Thanks!

[utkarshagarwal]

@pindinagesh
Yeah, I tried the installation guide but failed as I am using M1 Mac. Linux x86 VM on UTM also fails as it tries to leverage instruction set not supported by base CPU.

Any idea if these libs will be supported on new M1 Macs?

[parnell]

Any update on this issue? It would be nice to be able to use TF data validation on the M1. Especially as this hardware is becoming more and more present as older models get phased out.

[gaikwadrahul8]

@parnell, @utkarshagarwal

I'm really sorry to inform you that, at the moment unfortunately we don't support TFT, TFX and TFDV for M1 Mac, We are currently working on this issue, and will have an update in the fairly near future. In the meantime, some users have reported success with Rosetta. Other options include using a VM. We understand that neither of those is ideal

You can also check on our Tensorflow Forum, same discussion is going on here so you can check latest update about it.

Supported platforms :

TFDV is tested on the following 64-bit operating systems:

macOS 10.14.6 (Mojave) or later.
Ubuntu 16.04 or later.
Windows 7 or later.

[parnell]

Thank you for the update @gaikwadrahul8. Disappointing of course but I understand. Checking the discussion on the Tensorflow Forum it seems they recognize this as a real issue and that it is already impacting some Alphabet teams. So hopefully TFT, TFX, and TFDV will get supported soon. I'll be eagerly waiting.

[gaikwadrahul8]

@parnell

We completely understand Mac M1 users situation for TFT, TFX, and TFDV support and We are currently working on this issue, and will have an update in the fairly near future

Thank you!

[tangm]

Have created #251 . If/when tensorflow/tfx-bsl#73 goes through, it should allow tfdv to be built.

[ee07dazn]

Any update on where we are with this 🙏 ? Our entire team (company) is on Mac M1s so finding it extremely difficult to build something using TFDV.

[thesuperzapper]

If possible, we should also enable building for ARM64 on Linux.

[MohamadOHijazi]

If possible, we should also enable building for ARM64 on Linux.

Actually, This makes sense when You use TFX and tfdv in dataset preparation for model training you need to apply the operations on the new data aka inference data. and Linux and ARM64 are the perfect lightweight OS for an edge machine.
at least they should release a light version.