keras via tensorflow :: 2.13 :: AttributeError: module 'tensorflow.keras' has no attribute '__version__'

[bschulth]

When using newer tensorflow python packages (2.13, probably also 2.12), loading keras package fails with:
AttributeError: module 'tensorflow.keras' has no attribute '__version__'

Granted...the current (https://packagemanager.posit.co/cran/__linux__/jammy/2023-07-12/) version of
R keras is 2.11.1, so this might just be a compatibility issue between
R keras 2.11.1 and Python tensorflow/keras > 2.11.1

From R:

> keras:::keras_version()

2023-07-13 09:30:47.605075: I tensorflow/tsl/cuda/cudart_stub.cc:28] Could not find cuda drivers on your machine, GPU will not be used.
2023-07-13 09:30:47.656239: I tensorflow/tsl/cuda/cudart_stub.cc:28] Could not find cuda drivers on your machine, GPU will not be used.
2023-07-13 09:30:47.656834: I tensorflow/core/platform/cpu_feature_guard.cc:182] This TensorFlow binary is optimized to use available CPU instructions in performance-critical operations.
To enable the following instructions: AVX2 AVX512F FMA, in other operations, rebuild TensorFlow with the appropriate compiler flags.
2023-07-13 09:30:48.617657: W tensorflow/compiler/tf2tensorrt/utils/py_utils.cc:38] TF-TRT Warning: Could not find TensorRT
Error in py_get_attr_impl(x, name, silent) :
  AttributeError: module 'tensorflow.keras' has no attribute '__version__'
Run `reticulate::py_last_error()` for details.


> reticulate::py_last_error()

── Python Exception Message ────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────
AttributeError: module 'tensorflow.keras' has no attribute '__version__'

── R Traceback ─────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────
     ▆
  1. └─keras:::keras_version()
  2.   ├─keras$"__version__" at keras/R/package.R:279:2
  3.   └─reticulate:::`$.python.builtin.module`(keras, "__version__") at keras/R/package.R:279:2
  4.     └─reticulate:::py_resolve_module_proxy(x) at reticulate/R/python.R:296:2
  5.       └─keras (local) on_load() at reticulate/R/python.R:1527:2
  6.         └─keras:::keras_version() at keras/R/package.R:123:6
  7.           ├─keras$"__version__" at keras/R/package.R:279:2
  8.           └─reticulate:::`$.python.builtin.module`(keras, "__version__") at keras/R/package.R:279:2
  9.             └─reticulate:::`$.python.builtin.object`(x, name) at reticulate/R/python.R:299:2
 10.               └─reticulate:::py_get_attr_or_item(x, name, TRUE) at reticulate/R/python.R:407:2
 11.                 └─reticulate::py_get_attr(x, name) at reticulate/R/python.R:375:4
 12.                   └─reticulate:::py_get_attr_impl(x, name, silent) at reticulate/R/python.R:971:2

> packageVersion("keras")

[1] ‘2.11.1’

> packageVersion("tensorflow")

[1] ‘2.11.0’


> reticulate::py_version()
[1] ‘3.8’

> system("pip list | grep -E '(keras|tensorflow)'")

keras                             2.13.1
tensorflow                        2.13.0
tensorflow-cpu                    2.13.0
tensorflow-estimator              2.13.0
tensorflow-io-gcs-filesystem      0.32.0



> getOption("repos")
                                                              CRAN
"https://packagemanager.posit.co/cran/__linux__/jammy/2023-07-12/"



> sessionInfo()

R version 4.2.3 (2023-03-15)
Platform: x86_64-pc-linux-gnu (64-bit)
Running under: Ubuntu 22.04.2 LTS

Matrix products: default
BLAS/LAPACK: /opt/intel/compilers_and_libraries_2020.4.304/linux/mkl/lib/intel64_lin/libmkl_gf_lp64.so

locale:
 [1] LC_CTYPE=en_US.UTF-8       LC_NUMERIC=C
 [3] LC_TIME=en_US.UTF-8        LC_COLLATE=en_US.UTF-8
 [5] LC_MONETARY=en_US.UTF-8    LC_MESSAGES=en_US.UTF-8
 [7] LC_PAPER=en_US.UTF-8       LC_NAME=C
 [9] LC_ADDRESS=C               LC_TELEPHONE=C
[11] LC_MEASUREMENT=en_US.UTF-8 LC_IDENTIFICATION=C

attached base packages:
[1] stats     graphics  grDevices utils     datasets  methods   base

other attached packages:
[1] keras_2.11.1    testthat_3.1.10

loaded via a namespace (and not attached):
 [1] Rcpp_1.0.11       lattice_0.21-8    png_0.1-8         zeallot_0.1.0
 [5] rprojroot_2.0.3   brio_1.1.3        grid_4.2.3        R6_2.5.1
 [9] jsonlite_1.8.7    magrittr_2.0.3    tfruns_1.5.1      rlang_1.1.1
[13] cli_3.6.1         whisker_0.4.1     Matrix_1.6-0      reticulate_1.30
[17] generics_0.1.3    desc_1.4.2        pkgload_1.3.2.1   parallel_4.2.3
[21] compiler_4.2.3    base64enc_0.1-3   tensorflow_2.11.0

From Python:

# python
Python 3.8.3 (default, Jul 12 2023, 18:47:07)
[GCC 11.3.0] on linux
Type "help", "copyright", "credits" or "license" for more information.

>>> import tensorflow

2023-07-13 09:48:54.072448: I tensorflow/tsl/cuda/cudart_stub.cc:28] Could not find cuda drivers on your machine, GPU will not be used.
2023-07-13 09:48:54.119979: I tensorflow/tsl/cuda/cudart_stub.cc:28] Could not find cuda drivers on your machine, GPU will not be used.
2023-07-13 09:48:54.120575: I tensorflow/core/platform/cpu_feature_guard.cc:182] This TensorFlow binary is optimized to use available CPU instructions in performance-critical operations.
To enable the following instructions: AVX2 AVX512F FMA, in other operations, rebuild TensorFlow with the appropriate compiler flags.
2023-07-13 09:48:55.179887: W tensorflow/compiler/tf2tensorrt/utils/py_utils.cc:38] TF-TRT Warning: Could not find TensorRT

>>> tensorflow.__version__

'2.13.0'

>>> tensorflow.keras

<module 'keras.api._v2.keras' from '/opt/cfdads/python/3.8.3/lib/python3.8/site-packages/keras/api/_v2/keras/__init__.py'>

>>> tensorflow.keras.__version__

Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/opt/cfdads/python/3.8.3/lib/python3.8/site-packages/tensorflow/python/util/lazy_loader.py", line 59, in __getattr__
    return getattr(module, item)
AttributeError: module 'keras.api._v2.keras' has no attribute '__version__'


>>> import keras
>>> keras

<module 'keras' from '/opt/cfdads/python/3.8.3/lib/python3.8/site-packages/keras/__init__.py'>

>>> keras.__version__

'2.13.1'

[bschulth]

Just an update, I can confirm that downgrading my python install of tensorflow/keras to 2.11.1 resolves the issue.
So maybe this is just an FYI for next release of R keras to support 2.12+ python versions.

From R:

> keras:::keras_version()
2023-07-13 10:18:56.477483: I tensorflow/core/platform/cpu_feature_guard.cc:193] This TensorFlow binary is optimized with oneAPI Deep Neural Network Library (oneDNN) to use the following CPU instructions in performance-critical operations:  AVX2 AVX512F FMA
To enable them in other operations, rebuild TensorFlow with the appropriate compiler flags.
[1] ‘2.11.0’

> system("pip list | grep -E '(keras|tensorflow)'")

keras                             2.11.0
tensorflow                        2.11.1
tensorflow-cpu                    2.11.1
tensorflow-estimator              2.11.0
tensorflow-io-gcs-filesystem      0.32.0


> packageVersion("keras")

[1] ‘2.11.1’

> packageVersion("tensorflow")

[1] ‘2.11.0’

[t-kalinowski]

Can you please try with the development version of the R package?

remotes::install_github("rstudio/keras")

[bschulth]

Perfect! Thanks!
I can confirm the development version works with 2.13.0

> keras:::keras_version()
2023-07-13 12:08:01.417687: I tensorflow/tsl/cuda/cudart_stub.cc:28] Could not find cuda drivers on your machine, GPU will not be used.
2023-07-13 12:08:01.493816: I tensorflow/tsl/cuda/cudart_stub.cc:28] Could not find cuda drivers on your machine, GPU will not be used.
2023-07-13 12:08:01.495103: I tensorflow/core/platform/cpu_feature_guard.cc:182] This TensorFlow binary is optimized to use available CPU instructions in performance-critical operations.
To enable the following instructions: AVX2 AVX512F FMA, in other operations, rebuild TensorFlow with the appropriate compiler flags.
2023-07-13 12:08:02.784507: W tensorflow/compiler/tf2tensorrt/utils/py_utils.cc:38] TF-TRT Warning: Could not find TensorRT
[1] ‘2.13.0’


> system("pip list | grep -E '(keras|tensorflow)'")
keras                             2.13.1
tensorflow                        2.13.0
tensorflow-cpu                    2.13.0
tensorflow-estimator              2.13.0
tensorflow-io-gcs-filesystem      0.32.0

> packageVersion("keras")

[1] ‘2.11.1.9000’

> packageVersion("reticulate")

[1] ‘1.30.9000’