I have been stuck on this for a while and never thought that the issue was that anaconda was still using the Intel architecture. Your explanation and solution worked perfectly. Thanks!
This video ends at installation of tensorflow-deps. After this i guess we also need tensorflow-macos and tensorflow-metal. I did that, however jupyter notebook still throws ModuleNotFoundError: No module named 'tensorflow' and on CLI it says: RuntimeError: module compiled against API version 0xf but this version of numpy is 0xe RuntimeError: module compiled against API version 0xf but this version of numpy is 0xe ImportError: numpy.core._multiarray_umath failed to import ImportError: numpy.core.umath failed to import TypeError: Unable to convert function return value to a Python type! The signature was () -> handle
If you have older than M1, here is the instructions (the last step is critical): Install Miniforge Miniforge conda create -n your_env python=3.8.8 conda activate your_env python -m pip install tensorflow SYSTEM_VERSION_COMPAT=0 pip install tensorflow-macos tensorflow-metal
Thank you for the great instructional video! When trying to augment image data using Keras preprocessing layers (by making the preprocessing layers part of your model), I've noticed that M1Max is very slow (slower than Google Colab) and when using larger dataset, kernel dies. Is that natural apple metal? Also, can we use mixed precision with apple metal? Thank you.
Also the only problem I had was uninstalling anaconda. I followed the official website instructions to uninstall. Downloaded the M1 version anaconda, but I keep getting errors saying anaconda was already installed. used this "sudo rm -rf /opt/anaconda3" and it worked...
Please I need help : I followed all the steps and when I want to check the platform in the 2cases; (MacAir M2 Silicon), when I type platform.platform >> why do I have this :
Hi~Jeff, when i tried to import tensorflow on macbook air M2 chip, i had a error that “ illegal hardware instruction “. Can you help me solve this problem ?
2022-12-15 18:18:34.445783: I tensorflow/core/common_runtime/pluggable_device/pluggable_device_factory.cc:306] Could not identify NUMA node of platform GPU ID 0, defaulting to 0. Your kernel may not have been built with NUMA support. 2022-12-15 18:18:34.445803: I tensorflow/core/common_runtime/pluggable_device/pluggable_device_factory.cc:272] Created TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:0 with 0 MB memory) -> physical PluggableDevice (device: 0, name: METAL, pci bus id: ) when running tf then after run into a NotFoundError when running 100 epochs anyone have advice to resolve?
Updated version for 2023: ua-cam.com/video/o4-bI_iZKPA/v-deo.html
I have been stuck on this for a while and never thought that the issue was that anaconda was still using the Intel architecture. Your explanation and solution worked perfectly. Thanks!
YOU ARE THE GREATEST. I've been trying to solve this issue for hours and you saved me. Thank you sir!!!
You are the man! It took me 4 days to install tensorflow until I saw this
holy what.. that did it. The explanation, the logic. goodness... Thank you sir. Happy to subscribe.
Thank you!!! I was going crazy on figuring out that error message!
it helped me a lot, saving a million years to find the answer! thank you a billion times! :D
My issue comes on the next step regarding the tensorflow-metal install, MacbookPro M3Max
Thanks Jeff!
This video ends at installation of tensorflow-deps. After this i guess we also need tensorflow-macos and tensorflow-metal. I did that, however jupyter notebook still throws
ModuleNotFoundError: No module named 'tensorflow'
and on CLI it says:
RuntimeError: module compiled against API version 0xf but this version of numpy is 0xe
RuntimeError: module compiled against API version 0xf but this version of numpy is 0xe
ImportError: numpy.core._multiarray_umath failed to import
ImportError: numpy.core.umath failed to import
TypeError: Unable to convert function return value to a Python type! The signature was
() -> handle
You need to upgrade your numpy with pip
@@anchewei It was not related to numpy, i got it to work using miniforge for arm64
If you have older than M1, here is the instructions (the last step is critical):
Install Miniforge
Miniforge
conda create -n your_env python=3.8.8
conda activate your_env
python -m pip install tensorflow
SYSTEM_VERSION_COMPAT=0 pip install tensorflow-macos tensorflow-metal
Thank you for the great instructional video!
When trying to augment image data using Keras preprocessing layers (by making the preprocessing layers part of your model), I've noticed that M1Max is very slow (slower than Google Colab) and when using larger dataset, kernel dies. Is that natural apple metal?
Also, can we use mixed precision with apple metal?
Thank you.
Also the only problem I had was uninstalling anaconda. I followed the official website instructions to uninstall. Downloaded the M1 version anaconda, but I keep getting errors saying anaconda was already installed. used this "sudo rm -rf /opt/anaconda3" and it worked...
THank you. Very important point!
Or simply switch to PyTorch when its M1 support is now official through anaconda?
Please I need help : I followed all the steps and when I want to check the platform in the 2cases; (MacAir M2 Silicon), when I type platform.platform >> why do I have this :
Hi Jeff,
I purchased my M1 MacBook Pro 14. please help me which one I should install Miniconda Or Miniforge
Thanks very much;)
You saved my life!!!
Hi~Jeff, when i tried to import tensorflow on macbook air M2 chip, i had a error that “ illegal hardware instruction “. Can you help me solve this problem ?
Thank you so much!!
Thank You very much
Thanks!
2022-12-15 18:18:34.445783: I tensorflow/core/common_runtime/pluggable_device/pluggable_device_factory.cc:306] Could not identify NUMA node of platform GPU ID 0, defaulting to 0. Your kernel may not have been built with NUMA support.
2022-12-15 18:18:34.445803: I tensorflow/core/common_runtime/pluggable_device/pluggable_device_factory.cc:272] Created TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:0 with 0 MB memory) -> physical PluggableDevice (device: 0, name: METAL, pci bus id: )
when running tf then after run into a
NotFoundError when running 100 epochs
anyone have advice to resolve?
Great!!!
Thanks so much!
Great!