ERROR: numpy-1.18.5-cp38-cp38-macosx_11_0_x86_64.whl is not a supported wheel on this platform.

[soerenbrockmann]

I was trying to install tensorflow macos by running the script according to the instructions:

/bin/bash -c "$(curl -fsSL https://raw.githubusercontent.com/apple/tensorflow_macos/master/scripts/download_and_install.sh)"

However, I got an error as shown below:

Installing and upgrading base packages.
Requirement already satisfied: pip in ./tensorflow_macos_venv/lib/python3.8/site-packages (20.3)
Requirement already satisfied: wheel in ./tensorflow_macos_venv/lib/python3.8/site-packages (0.35.1)
Requirement already satisfied: setuptools in ./tensorflow_macos_venv/lib/python3.8/site-packages (50.3.2)
Requirement already satisfied: cached-property in ./tensorflow_macos_venv/lib/python3.8/site-packages (1.5.2)
Requirement already satisfied: six in ./tensorflow_macos_venv/lib/python3.8/site-packages (1.15.0)

Installing bundled binary dependencies.
ERROR: numpy-1.18.5-cp38-cp38-macosx_11_0_x86_64.whl is not a supported wheel on this platform.

Any idea how to fix it?

[JLP04]

I also just started seeing this issue but I'm not sure how to fix it.

[kharteveld]

The problem seems to be that all the bundled wheel files are called macos_11_0, but pip thinks big sur is macOS_10_9. I renamed all the bundled wheel files to macos_10_9 and then it worked.

[ismaproco]

@kharteveld how did you the renaming? I would like to try it

[soerenbrockmann]

Alright, I found it.

1. Download the archive file manually as mention at the README file.
2. Unpack it and rename all files inside x86_64/ folder. They all contain _11_0. Rename it to _10_16

3. Open install_venv.sh and rename those filenames (macosx_11_0 to macosx_10_16)
   arch_list_x86_64=( numpy-1.18.5-cp38-cp38-macosx_10_16_x86_64.whl grpcio-1.33.2-cp38-cp38-macosx_10_16_x86_64.whl h5py-2.10.0-cp38-cp38-macosx_10_16_x86_64.whl scipy-1.5.4-cp38-cp38-macosx_10_16_x86_64.whl tensorflow_addons-0.11.2+mlcompute-cp38-cp38-macosx_10_16_x86_64.whl )

4. Scroll down and rename macosx_11_0 to macosx_10_16

"$python_bin" -m pip install --upgrade --force -t "$VIRTUAL_ENV/lib/python3.8/site-packages/" --no-dependencies "$package_dir"/tensorflow_macos*-cp38-cp38-macosx_10_16_$arch.whl

5. Open terminal, go inside your tensorflow_macos folder and run
   ./install_venv.sh /Users/<your user>/tensorflow_macos_venv/

Those steps worked for me. :-D

[rnogy]

I don't think you have to rename it; the trick is just to use -t hack. So you can do like

#!/bin/bash
arch_list_x86_64=( numpy-1.18.5-cp38-cp38-macosx_11_0_x86_64.whl
                   grpcio-1.33.2-cp38-cp38-macosx_11_0_x86_64.whl       
                   h5py-2.10.0-cp38-cp38-macosx_11_0_x86_64.whl
		   scipy-1.5.4-cp38-cp38-macosx_11_0_x86_64.whl  
                   tensorflow_addons-0.11.2+mlcompute-cp38-cp38-macosx_11_0_x86_64.whl
                   tensorflow_macos-0.1a0-cp38-cp38-macosx_11_0_x86_64.whl )

for f in ${arch_list_x86_64[@]} ; do 
"$(which python)" -m pip install --upgrade -t "$(pip show pip | grep Location | cut -c 11-)" --no-dependencies --force "$f"
done

(modified from apple's install_venv.sh script)
on whatever env you're in without using apple's script (which you don't have to install under venv).

[seibert]

Note that pip 20.3 was released yesterday and may have fixed some of these issues:

https://pip.pypa.io/en/stable/news/#id1

[monikavila]

Note that pip 20.3 was released yesterday and may have fixed some of these issues:

https://pip.pypa.io/en/stable/news/#id1

It does not solve the problem

[monikavila]

@soerenbrockmann did your solution work for the new MacBook Pro M1?
Also, I cannot download the files anymore. It seems that they are not available.

[godkaieethu]

@soerenbrockmann did your solution work for the new MacBook Pro M1?
Also, I cannot download the files anymore. It seems that they are not available.

I have tried, but it seems doesn't work for m1

[nuwansam]

I tried the same in a brand new M1 using python 3.8. I tried installing both architecture types and I get the "is not a supported wheel on this platform" for both (for numpy), which stops the installer.

[soerenbrockmann]

@monikavila I am working on a MacBookPro16 8-Core Intel Core i9 :-) I haven't tested it on M1. You can find the files here:

https://github.com/apple/tensorflow_macos/releases

However, I am facing lot's of Python dependency issue recently too. Not sure what's going on :-D

[monikavila]

@monikavila I am working on a MacBookPro16 8-Core Intel Core i9 :-) I haven't tested it on M1. You can find the files here:

https://github.com/apple/tensorflow_macos/releases

However, I am facing lot's of Python dependency issue recently too. Not sure what's going on :-D

Thanks a lot!

[monikavila]

I tried the same in a brand new M1 using python 3.8. I tried installing both architecture types and I get the "is not a supported wheel on this platform" for both (for numpy), which stops the installer.

I tried to update to Rosetta 2 and do the whole installation again but it does not work. I also tried to change to macosx_10_16 and I get

ERROR: numpy-1.18.5-cp38-cp38-macosx_10_16_arm64.whl is not a supported wheel on this platform.

I will contact apple support. If you contact apple support also, it would be good and please let me know.

[icenando]

I tried the same in a brand new M1 using python 3.8. I tried installing both architecture types and I get the "is not a supported wheel on this platform" for both (for numpy), which stops the installer.

I tried to update to Rosetta 2 and do the whole installation again but it does not work. I also tried to change to macosx_10_16 and I get

ERROR: numpy-1.18.5-cp38-cp38-macosx_10_16_arm64.whl is not a supported wheel on this platform.

I will contact apple support. If you contact apple support also, it would be good and please let me know.

Please do update us on how Apple responds to this. I have a feeling they will say it's not their problem and we should contact Google instead.

[icenando]

So this worked for me: #53 (comment) .

[icenando]

This (kind of) worked for me:

• Anaconda is already removed. I used Anaconda_clean.

• Removed all versions of python and related files from:

  • Applications
  • /Library/Frameworks/Python.framework/Versions
  • /usr/local/bin

• Uninstalled Xcode

• Restarted Mac

Checked Python version in Terminal (“python —version): Python 2.7.16 (which I’m guessing is the one that comes embedded in Big Sur.

• Reinstalled Xcode
• Ran Xcode-select --install from Terminal.
• Installed tensor flow again.

Except now I can’t seem to import anything else! I’m getting pages and pages of errors trying to pip install pandas and matplotlib...

[kukupigs]

1. ./install_venv.sh /Users//tensorflow_macos_venv/

it's work.good job!

[kabouzeid]

Just use --python=/usr/bin/python3 in install_venv.sh

[SaschaHeyer]

Is there any official update on that. Issue still exist and this ticket is now open for more than 1 month.

[Aizen741]

@SaschaHeyer @monikavila @icenando @soerenbrockmann
Dont create new Virtual Environment under Anaconda. You have to create it under Miniforge

First, get out of the anaconda env , and after you download the miniforge create the Virtual Environment under Miniforge .

Once you create Env in Miniforge this will come . So Don't Close this terminal yet. open a new terminal

Download Additional file

Downloading the Tensorflow : Download tensorflow from releases , Once you download and unzip it open the file and drag - drop install_venv into terminal under the env also add -p in the end. it will ask for the location . provide the same location as the env.

There You go . Tensorflow is working for me perfectly on m1 even in jupyter notebook too

[tinochan]

The solution by @Aizen741 works! It also works for Jupyter Notebook by creating a new kernel using the virtual environment under Miniforge.