this is my code
random = tf.random.Generator.from_seed(42)
random = random.normal(shape=(2,2))
but i got this error:
tensorflow.python.framework.errors_impl.NotFoundError: No registered 'RngReadAndSkip'
OpKernel for 'GPU' devices compatible with node {{node RngReadAndSkip}}. Registered: device='CPU'
It happens because some TF/Keras high-level methods are not implemented in tensorflow-metal. You can try this workaround:
with tf.device('/cpu:0'):
... # your code which need RngReadAndSkip low-level method here
And next you can do other stuff normally (with GPU).
Hope, Apple will fix it someday (and their another tons of bugs too, lmao)