Installation Issue for M1 Chip

[cadyyuheng]

Hello,

Although I have read several posts on handling the "TensorFlow not found" installation issue, I am still unable to successfully install Keras. I restarted R session and installed with commands:

remotes::install_github("rstudio/tensorflow")
remotes::install_github("rstudio/keras")
reticulate::miniconda_uninstall()
reticulate::install_miniconda()
keras::install_keras()

I can find and tensorflow-macro and tensorflow-metal in r-reticulate. Yet when I try to test keras installation, I got the following error:

> tensorflow::as_tensor("Hello World")
RuntimeError: module compiled against API version 0x10 but this version of numpy is 0xf
RuntimeError: module compiled against API version 0x10 but this version of numpy is 0xf
Error: Valid installation of TensorFlow not found.

Python environments searched for 'tensorflow' package:
 /Users//Library/r-miniconda-arm64/envs/r-reticulate/bin/python3.8

Python exception encountered:
 ImportError: numpy.core.multiarray failed to import

The above exception was the direct cause of the following exception:

SystemError: <built-in method __contains__ of dict object at 0x11f316bc0> returned a result with an error set

The above exception was the direct cause of the following exception:

Traceback (most recent call last):
  File "/Library/Frameworks/R.framework/Versions/4.2-arm64/Resources/library/reticulate/python/rpytools/loader.py", line 39, in _import_hook
    module = _import(
  File "/Users//Library/r-miniconda-arm64/envs/r-reticulate/lib/python3.8/site-packages/tensorflow/__init__.py", line 37, in <module>
    from tensorflow.python.tools import module_util as _module_util
  File "/Library/Frameworks/R.framework/Versions/4.2-arm64/Resources/library

Attached is my session information and I sincerely hope you could help me with this.

> reticulate::py_config()
python:         /Users//Library/r-miniconda-arm64/envs/r-reticulate/bin/python3.8
libpython:      /Users//Library/r-miniconda-arm64/envs/r-reticulate/lib/libpython3.8.dylib
pythonhome:     /Users//Library/r-miniconda-arm64/envs/r-reticulate:/Users//Library/r-miniconda-arm64/envs/r-reticulate
version:        3.8.13 | packaged by conda-forge | (default, Mar 25 2022, 06:05:16)  [Clang 12.0.1 ]
numpy:          /Users//Library/r-miniconda-arm64/envs/r-reticulate/lib/python3.8/site-packages/numpy
numpy_version:  1.22.4
tensorflow:     /Users//Library/r-miniconda-arm64/envs/r-reticulate/lib/python3.8/site-packages/tensorflow

NOTE: Python version was forced by use_python function
> tensorflow::tf_config()
RuntimeError: module compiled against API version 0x10 but this version of numpy is 0xf
RuntimeError: module compiled against API version 0x10 but this version of numpy is 0xf
Valid installation of TensorFlow not found.

Python environments searched for 'tensorflow' package:
 /Users//Library/r-miniconda-arm64/envs/r-reticulate/bin/python3.8

Python exception encountered:
 ImportError: numpy.core.multiarray failed to import

The above exception was the direct cause of the following exception:

SystemError: <built-in method __contains__ of dict object at 0x11f35ed80> returned a result with an error set

The above exception was the direct cause of the following exception:

Traceback (most recent call last):
  File "/Library/Frameworks/R.framework/Versions/4.2-arm64/Resources/library/reticulate/python/rpytools/loader.py", line 39, in _import_hook
    module = _import(
  File "/Users//Library/r-miniconda-arm64/envs/r-reticulate/lib/python3.8/site-packages/tensorflow/__init__.py", line 37, in <module>
    from tensorflow.python.tools import module_util as _module_util
  File "/Library/Frameworks/R.framework/Versions/4.2-arm64/Resources/library/reticulate/python/rpytools/loader.py", line 39, in _import_hook
    module = _import(
  File "/Users//Library/r-miniconda-arm64/envs/r-reticulate/lib/python3.8/site-packages/tensorflow/python/__init__.py", line 37, in <module>
    from tensorflow.python.eager import context
  File "/Library/Frameworks/R.framework/Versions/4.2-arm64/Resources/library/reticulate/python/rpytools/loader.py", line 39, in _import_hook
    module = _import(
  File "/Users//Library/r-miniconda-arm64/envs/r-reticulate/lib/python3.8/site-packages/tensorflow/python/eager/context.py", line 35, in <module>
    from tensorflow.python.client import pywrap_tf_session
  File "/Library/Frameworks/R.framework/Versions/4.2-arm64/Resources/library/reticulate/python/rpytools/loader.py", line 39, in _import_hook
    module = _import(
  File "/Users//Library/r-miniconda-arm64/envs/r-reticulate/lib/python3.8/site-packages/tensorflow/python/client/pywrap_tf_session.py", line 19, in <module>
    from tensorflow.python.client._pywrap_tf_session import *
  File "/Library/Frameworks/R.framework/Versions/4.2-arm64/Resources/library/reticulate/python/rpytools/loader.py", line 39, in _import_hook
    module = _import(
ImportError: initialization failed


You can install TensorFlow using the install_tensorflow() function.
 
> reticulate::import("tensorflow")
RuntimeError: module compiled against API version 0x10 but this version of numpy is 0xf
Error in py_module_import(module, convert = convert) : 
  ImportError: initialization failed

[cadyyuheng]

This is now resolved. After running

reticulate::miniconda_uninstall()
reticulate::install_miniconda()

The numpy version for the r-reticulate environment is 1.23.3. Yet after running

keras::install_keras()

The numpy version was downgraded to 1.22.4, which I suspect caused the error message running tensorflow::as_tensor("Hello World")

Python exception encountered:
 ImportError: numpy.core.multiarray failed to import

After I manually upgraded the numpy back to 1.23.3in r-reticulate, tensorflow::as_tensor("Hello World") gave successful output.
I wonder if the keras team could solve this issue for future updates.