I’ve set up PyCharm on my M1 machine as per the exam handbook’s instructions.
However, running the script below:
import tensorflow as tf
print(tf.version)
Gives this error: Process finished with exit code 132 (interrupted by signal 4: SIGILL)
From my google search, the exam’s version of TensorFlow seems incompatible with my machine.
But the handbook was very strict about the tensorflow version - otherwise the grader won’t work.
How do I resolve this?
To offer context, my macOS is version 11.4 with the M1 chip.
I’ve tried both intel and the apple silicon versions of Pycharm Community.
The command “python3 --version” successfully shows python 3.8.6 (3.8.x is required for exam)
I’ve tried installing the apple tensorflow for 11.x os versions.
Update: It is finally running on my M1. Please install the release from apple/tensorflow_macros. Also, make sure that you run the script in the right arch i.e. arm64
Does anyone have any updates on this? I’m able to use tensorflow-macos with python 3.8 on my M1 Macbook Pro, but the standard tensorflow package is still a no go. Am I going to have to find a different machine to use for the exam?
Sorry but this doesn’t really help. The linked instructions on developer.apple.com advise installing the tensforflow-macos package, which I have successfully done (and it works fine).
The documentation I have read for the certification exam indicates that the testing environment will use the vanilla tensorflow package, but there are no such packages for aarch64 Macs on PyPi (only x86_64).
I sent in a question about this to the TF exam support and this was the answer they gave me.
“”"
We have not tested the TensorFlow Developer Certificate exam on a computer with M1.
The exam plugin will try to install TensorFlow version 2.7.
The exam will only work with the standard implementation of TensorFlow, not with tensorflow-macos.
“”"
From that response it looks like the exam configuration will fail on any newer mac that doesn’t have the intel architecture. I think it’s a bad practice to exclude such a large number of developers from taking this exam (and growing as people update or buy a new computer) but that might be the case at the moment.
If anyone has figured out some way around this it would be great to know.
I upgraded my computer from an old Windows PC to a new M2 Mac Mini so that I could escape dependency-hell. (I have to admit that getting Tensorflow running on this thing was smooth as can be though.)
It’s a wee-bit depressing that I went through months of courses to prepare for and take this exam just in order to find out that incompatible packages would get me once again.
Tagging this post so that hopefully soon (ideally within 6 months) they update the exam to and I can check back on the good news.
I sent an email there a few days ago with no response so far.
Since then I’ve spent a dozen hours trying to get my old Windows computer set up with the right old CUDA & cuDNN libraries and ended up just completely reinstalling Windows just to hopefully have an easier time getting TF2.9 running to take the cert while it’s all still fresh in my head.
@chunduriv
Any update on attempting tensorflow certificate on mac m1. Can you please provide a setup guide for tensorflow developer exam specifically for Macbook m1.
Thanks in advance
Hi @Kiran_Sai_Ramineni I went through the stack overflow solution but the installation works with python 3.10 but as per Tensorflow setup guide it required python 3.8. I tried with python 3.8 but tensorflow doesn’t install with this version of python. It gives the below error.
ERROR: Could not find a version that satisfies the requirement tensorflow-macos==2.9 (from versions: none).
Can you please help me setup my mac m1 as per specifications for tensorflow developer certificate examination.
Thanks in advance
Does anybody tried a workaround for this issue? e.g. to use a virtual machine (with emulation of windows or linux) and setting up the environment in the virtual machine?
I have recently bought a macbook pro M2 for developing ML and I have just read this thread, I was not aware about this issue and it is really frustrating.
