Keras: Error in py_call_impl(callable, dots$args, dots$keywords) : AttributeError: 'int' object has no attribute 'ndim'
Hello,
This issue is very similar to https://github.com/rstudio/keras/issues/33 but with a different error. I can't run any of the R keras imdb examples but can run the mnist examples.
This is a sample of one of the imdb examples that won't run.
library(keras)
use_condaenv('py35', required = TRUE)
reticulate::py_config()
tf_config()
# Parameters --------------------------------------------------------------
# Embedding
max_features = 20000
maxlen = 100
embedding_size = 128
# Convolution
kernel_size = 5
filters = 64
pool_size = 4
# LSTM
lstm_output_size = 70
# Training
batch_size = 30
epochs = 2
# Data Preparation --------------------------------------------------------
# The x data includes integer sequences, each integer is a word
# The y data includes a set of integer labels (0 or 1)
# The num_words argument indicates that only the max_fetures most frequent
# words will be integerized. All other will be ignored.
# See help(dataset_imdb)
imdb <- dataset_imdb(num_words = max_features)
# Keras load all data into a list with the following structure:
str(imdb)
# Pad the sequences to the same length
# This will convert our dataset into a matrix: each line is a review
# and each column a word on the sequence
# We pad the sequences with 0s to the left
x_train <- imdb$train$x %>%
pad_sequences(maxlen = maxlen)
x_test <- imdb$test$x %>%
pad_sequences(maxlen = maxlen)
# Defining Model ------------------------------------------------------
model <- keras_model_sequential()
model %>%
layer_embedding(max_features, embedding_size, input_length = maxlen) %>%
layer_dropout(0.25) %>%
layer_conv_1d(
filters,
kernel_size,
padding = "valid",
activation = "relu",
strides = 1
) %>%
layer_max_pooling_1d(pool_size) %>%
layer_lstm(lstm_output_size) %>%
layer_dense(1) %>%
layer_activation("sigmoid")
model %>% compile(
loss = "binary_crossentropy",
optimizer = "adam",
metrics = "accuracy"
)
# Training ----------------------------------------------------------------
model %>% fit(
x_train, imdb$train$y,
batch_size = batch_size,
epochs = epochs,
validation_data = list(x_test, imdb$test$y)
)
As in the other ticket I saw this is the output i get from reticulate
> reticulate::py_config()
python: C:\Users\jmauzey\AppData\Local\conda\conda\envs\r-tensorflow\python.exe
libpython: C:/Users/jmauzey/AppData/Local/conda/conda/envs/r-tensorflow/python36.dll
pythonhome: C:\Users\jmauzey\AppData\Local\conda\conda\envs\R-TENS~1
version: 3.6.4 |Anaconda, Inc.| (default, Jan 16 2018, 10:22:32) [MSC v.1900 64 bit (AMD64)]
Architecture: 64bit
numpy: C:\Users\jmauzey\AppData\Local\conda\conda\envs\R-TENS~1\lib\site-packages\numpy
numpy_version: 1.14.0
keras: C:\Users\jmauzey\AppData\Local\conda\conda\envs\R-TENS~1\lib\site-packages\keras
And the output from the python versions found
python versions found:
C:\Users\jmauzey\AppData\Local\conda\conda\envs\r-tensorflow\python.exe
C:\Users\jmauzey\AppData\Local\CONTIN~1\ANACON~1\python.exe
C:\Users\jmauzey\AppData\Local\Continuum\anaconda3\python.exe
C:\Users\jmauzey\AppData\Local\conda\conda\envs\my_root\python.exe
C:\Users\jmauzey\AppData\Local\conda\conda\envs\tf-keras\python.exe
I have been struggling with this for a couple of days. I had been experiencing another issue and read that uninstalling anaconda would help. I removed it and it solved my problem. I used R keras for a week but then needed jupyter notebook again so reinstalled anaconda thinking that a fresh installation would help. R keras hasn't worked after that. I have attempted uninstalling R, R studio, all python and anaconda installations in an attempt to start from scratch. I have reinstalled reticulate as well. I have also noticed this message when using install_keras()
> install_keras()
Using r-tensorflow conda environment for TensorFlow installation
Solving environment: ...working...
WARNING: The remote server could not find the noarch directory for the
requested channel with url: https://conda.anaconda.org/conda-forge
It is possible you have given conda an invalid channel. Please double-check
your conda configuration using `conda config --show`.
If the requested url is in fact a valid conda channel, please request that the
channel administrator create `noarch/repodata.json` and associated
`noarch/repodata.json.bz2` files, even if `noarch/repodata.json` is empty.
$ mkdir noarch
$ echo '{}' > noarch/repodata.json
$ bzip2 -k noarch/repodata.json
done
# All requested packages already installed.
Here are some additional messages I got when using the same commands I saw in the other ticket.
> tf_config()
TensorFlow v1.2.1 (C:\Users\jmauzey\AppData\Local\conda\conda\envs\R-TENS~1\lib\site-packages\keras)
Python v3.6 (C:\Users\jmauzey\AppData\Local\conda\conda\envs\r-tensorflow\python.exe)
> devtools::session_info()
Session info ----------------------------------------------------------------------------------------------------------------------------------------------
setting value
version R version 3.4.3 (2017-11-30)
system x86_64, mingw32
ui RStudio (1.1.419)
language (EN)
collate English_United States.1252
tz America/Denver
date 2018-01-26
Packages --------------------------------------------------------------------------------------------------------------------------------------------------
package * version date source
base * 3.4.3 2017-12-06 local
base64enc 0.1-3 2015-07-28 CRAN (R 3.4.0)
compiler 3.4.3 2017-12-06 local
datasets * 3.4.3 2017-12-06 local
devtools 1.13.4 2017-11-09 CRAN (R 3.4.3)
digest 0.6.14 2018-01-14 CRAN (R 3.4.3)
graphics * 3.4.3 2017-12-06 local
grDevices * 3.4.3 2017-12-06 local
jsonlite 1.5 2017-06-01 CRAN (R 3.4.3)
keras * 2.1.3 2018-01-17 CRAN (R 3.4.3)
magrittr 1.5 2014-11-22 CRAN (R 3.4.0)
memoise 1.1.0 2017-04-21 CRAN (R 3.4.0)
methods * 3.4.3 2017-12-06 local
R6 2.2.2 2017-06-17 CRAN (R 3.4.3)
Rcpp 0.12.15 2018-01-20 CRAN (R 3.4.3)
reticulate 1.4 2018-01-09 CRAN (R 3.4.3)
stats * 3.4.3 2017-12-06 local
tensorflow * 1.5 2018-01-17 CRAN (R 3.4.3)
tfruns 1.2 2018-01-18 CRAN (R 3.4.3)
tools 3.4.3 2017-12-06 local
utils * 3.4.3 2017-12-06 local
whisker 0.3-2 2013-04-28 CRAN (R 3.4.0)
withr 2.1.1 2017-12-19 CRAN (R 3.4.3)
yaml 2.1.16 2017-12-12 CRAN (R 3.4.3)
zeallot 0.0.6 2017-09-28 CRAN (R 3.4.3)
Let me know anything else I can provide.
mauzeyj
All 13 comments
I believe I am having problems with my numpy installation. Sorry about bothering. I am hoping to get it fixed. I am not sure how to close a ticket I created.
mauzeyj
on 26 Jan 2018
I do get exactly the same error while running the IMDB example from above.
Error in py_call_impl(callable, dots$args, dots$keywords) :
AttributeError: 'int' object has no attribute 'ndim'
Here is the output of py_config:
reticulate::py_config()
#> python: /Users/LVG/miniconda3/envs/r-tensorflow/bin/python
#> libpython: /Users/LVG/miniconda3/envs/r-tensorflow/lib/libpython3.6m.dylib
#> pythonhome: /Users/LVG/miniconda3/envs/r-tensorflow:/Users/LVG/miniconda3/envs/r-tensorflow
#> version: 3.6.4 |Anaconda, Inc.| (default, Jan 16 2018, 12:04:33) [GCC 4.2.1 Compatible Clang 4.0.1 (tags/RELEASE_401/final)]
#> numpy: /Users/LVG/miniconda3/envs/r-tensorflow/lib/python3.6/site-packages/numpy
#> numpy_version: 1.14.0
#>
#> python versions found:
#> /Users/LVG/miniconda3/envs/r-tensorflow/bin/python
#> /usr/bin/python
#> /Users/LVG/miniconda3/bin/python
Any help will be much appreciated.
Thank you.
devtools::session_info()
#> Session info -------------------------------------------------------------
#> setting value
#> version R version 3.4.3 (2017-11-30)
#> system x86_64, darwin15.6.0
#> ui X11
#> language (EN)
#> collate en_CA.UTF-8
#> tz America/Montreal
#> date 2018-01-28
#> Packages -----------------------------------------------------------------
#> package * version date source
#> backports 1.1.2 2017-12-13 CRAN (R 3.4.3)
#> base * 3.4.3 2017-12-07 local
#> compiler 3.4.3 2017-12-07 local
#> datasets * 3.4.3 2017-12-07 local
#> devtools 1.13.4 2017-11-09 CRAN (R 3.4.2)
#> digest 0.6.14 2018-01-14 CRAN (R 3.4.3)
#> evaluate 0.10.1 2017-06-24 CRAN (R 3.4.1)
#> graphics * 3.4.3 2017-12-07 local
#> grDevices * 3.4.3 2017-12-07 local
#> htmltools 0.3.6 2017-04-28 CRAN (R 3.4.0)
#> knitr 1.18 2017-12-27 CRAN (R 3.4.3)
#> magrittr 1.5 2014-11-22 CRAN (R 3.4.0)
#> memoise 1.1.0 2017-04-21 CRAN (R 3.4.0)
#> methods * 3.4.3 2017-12-07 local
#> Rcpp 0.12.15 2018-01-20 CRAN (R 3.4.3)
#> rmarkdown 1.8 2017-11-17 CRAN (R 3.4.2)
#> rprojroot 1.3-2 2018-01-03 CRAN (R 3.4.3)
#> stats * 3.4.3 2017-12-07 local
#> stringi 1.1.6 2017-11-17 CRAN (R 3.4.2)
#> stringr 1.2.0 2017-02-18 CRAN (R 3.4.0)
#> tools 3.4.3 2017-12-07 local
#> utils * 3.4.3 2017-12-07 local
#> withr 2.1.1 2017-12-19 CRAN (R 3.4.3)
#> yaml 2.1.16 2017-12-12 CRAN (R 3.4.3)
LVG77
on 29 Jan 2018
@jpip Could you please tell me how you fixed that problem.
Thanks.
ulsmanikanta
on 29 Jan 2018
A few more observations about this error.
- I am not able to reproduce it on a Windows machine. However, I have been able to reproduce it on 3 different Mac OS machines (2 with anaconda and 1 without).
- The issue is somehow linked to the
validation_data = list(x_test, imdb$test$y)argument of thefitfunction. If this argument is removed, the problem disappears. If we usevalidation_splitargument instead, the problem disappears again.
LVG77
on 29 Jan 2018
I actually have not been able to resolve this issue yet. Just did what @LVG77 recommended and my model is now training. This is odd. The mnist examples use that same line of code.
mauzeyj
on 29 Jan 2018
I use validation split instead it works as @LVG77 said.
mauzeyj
on 30 Jan 2018
Fixed here: https://github.com/rstudio/keras/commit/efae072c7f18f6910a25c804176de0692d29c6cb
This was indeed an issue introduced in Keras 2.1.3 which is why I couldn't repro it at first. Thanks for pointing out the source of the trouble!
jjallaire
on 2 Feb 2018
In case anyone is following along in "Deep Learning with R" and getting this error on the first IMDB example: in the fit step, just change the line
history <- model %>% fit(
partial_x_train,
partial_y_train,
epochs = 20,
batch_size = 512,
validation_data = list(x_val, y_val)
)
to
history <- model %>% fit(
x_train,
y_train,
epochs = 20,
batch_size = 512,
validation_split = 0.4
)
mpelath
on 12 Feb 2018
This should also be fixed in the development version of the keras R package, which you can install with:
devtools::install_github("rstudio/keras")
I still get the same error: "Error in py_call_impl(callable, dots$args, dots$keywords)".
I did install the development version from github, but when I try to run the command
history <- model %>% fit_generator(
train_gen,
steps_per_epoch = 500,
epochs = 20,
validation_data = val_gen,
validation_steps = val_steps
)
I get this error.
miladm12
on 18 Apr 2018
Had the same problem in 2019 (following Deep learning with R book) and installing the dev version from github solved the problem.
romunov
on 21 Aug 2019
Had the same problem in 2019 (following Deep learning with R book) and installing the dev version from github solved the problem.
Hi, I am too following the Deep learning R book and I did install keras from devtools, but I keep getting this error when I run install_keras()
Error 1 occured creating conda enviroment r-reticulate.
How can I change the PATH to be the correct one?
mousa-saeed
on 4 Nov 2019
@saikmsm I suggest you open a new issue pertinent to your error message.
romunov
on 4 Nov 2019
Related issues
fderyckel
路
4Comments
faltinl
路
6Comments
MaximilianPi
路
5Comments
gsimchoni
路
3Comments
jingweihuang
路
6Comments
Most helpful comment
This should also be fixed in the development version of the keras R package, which you can install with: