I want to use the tfx 1.6.1 on Mac(M1 Max)

I want to use the tfx 1.6.1 version in my MacBook M1 Max, but it doesn’t seem compatible. Is there a way to solve it?


We are currently working on this issue, and will have an update in the fairly near future. Probably weeks, rather than months. In the meantime, some users have reported success with Rosetta. Other options include using a VM. We understand that neither of those is ideal.


Bumping along with tensorflow-serving-api availability for M1s

@Robert_Crowe HI, Do you have any update for apple M1.

Unfortunately this has taken longer than expected. I don’t have a firm ETA yet, but I know that folks are working on it.

Hi @Robert_Crowe , unfortunately, I am also waiting for this M1-compatible version, because I would like to use the module for my university project. Is there any news about the development status?
kind regards

I’m afraid that currently the best answer is to use Rosetta, which I know is not ideal. I’m trying to raise the priority for this, but there are internal infrastructure reasons for why this has proceeded slowly.

BTW, this is also a problem already (and growing) for several Alphabet teams, who also have M1 laptops.

Yes I will try using it with rosetta…but thanks a lot for trying to raise the priority. I am looking forward to try it when it is finished for M1!

1 Like

Hi @Robert_Crowe!

As far as I know, using Rosetta doesn’t work for TF(X) libraries since Rosetta can’t emulate AVX instructions.

Our team is also interested in running TFX in an linux/aarch64 environment (not specifically for macOS, i.e. a manylinux2014_aarch64 wheel rather than macosx_11_0_arm64 wheel).

Is there any ETA for aarch64 TFX? Since this was recently done for core TF, I assume the path forward is relatively straightforward?


I would also like to use tfx==1.10.0 tensorflow-macos==2.10.0 tensorflow-metal==0.6.0 on apple silicon arm64

Hey @Robert_Crowe - I know you from the mlops specialization - your’e great!
I’m also pumping this issue
I want to use TFX - first on my machine on docker compose before going to the cloud.
I tried with Rosetta and it failed to on AVX instructions
Is there any progress with it?