I am attempting to run an app using Tensorflow on the M1 platform.
I created my new conda environment, activated it, installed my other dependencies, and followed this guide from Apple: https://developer.apple.com/metal/tensorflow-plugin/
installing the tensorflow-metal frameworks at the end as specified. Then, I ran my python code, and crashed on
from keras.models import load_model
with
Symbol not found: OBJC_CLASS$_MPSGraphCompilationDescriptor
referenced from
libmetal_plugin.dylib
expected in
MetalPerformanceShadersGraph
Any ideas what I am doing wrong? I am running on an M1 Mac mini running macOS 11.6.1. Thank you.