Prerequisites for installing TensorFlow not available - r

I am trying to install Tensorflow in R as explained here- https://github.com/rstudio/tensorflow. I want to use Keras for deep learning in R.
devtools::install_github("rstudio/tensorflow")
library(tensorflow)
install_tensorflow()
but there was an error
Error: Prerequisites for installing TensorFlow not available.
Execute the following at a terminal to install the prerequisites:
$ sudo /usr/local/bin/pip install --upgrade virtualenv
i followed the code given to put into terminal but this seems like another error
The directory '/Users/liangweiteo/Library/Caches/pip/http' or its
parent directory is not owned by the current user and the cache has
been disabled. Please check the permissions and owner of that
directory. If executing pip with sudo, you may want sudo's -H flag.
The directory '/Users/liangweiteo/Library/Caches/pip' or its parent
directory is not owned by the current user and caching wheels has been
disabled. check the permissions and owner of that directory. If
executing pip with sudo, you may want sudo's -H flag. Requirement
already up-to-date: virtualenv in
/Library/Frameworks/Python.framework/Versions/2.7/lib/python2.7/site-packages
tried to look for help on this site
Error while installing Tensorflow in R
but somehow the error is still there

Related

RStudio Server (community) can't find libR.so even though it is in path

Note: I have filed this as an issue, but I'm not sure if it is really a bug or just something I need to resolve about my system configuration.
This seems to be a library that causes many people trouble with RStudio and RStudio Server. Often people can fix the problem by reinstalling the core R libraries with apt or manually copying or linking the libR.so file to a place where RStudio finds it.
In my case, I'm using a Conda instance for my R executable.
My instance was working and stopped after upgrading my ubuntu 22.04 VM. I tried some things to fix the problem but have not succeeded.
System details
RStudio Edition : Server
RStudio Version : 2022.07.2+576 (Spotted Wakerobin) for Ubuntu Bionic
OS Version : Ubuntu 22.04.1 LTS
R Version : 4.1.3 (2022-03-10) -- "One Push-Up"
Describe the problem in detail
I have a GCP VM running ubuntu 22.04 which I use for RSS.
I did a sudo apt update && sudo apt dist-upgrade and a restart. RSS stopped working. I ran sudo rstudio-server verify-installation and received
/usr/lib/rstudio-server/bin/rsession: error while loading shared libraries: libR.so: cannot open shared object file: No such file or directory
I decided to reinstall RSS using:
wget https://download2.rstudio.org/server/jammy/amd64/rstudio-server-2022.07.2-576-amd64.deb
sudo gdebi rstudio-server-2022.07.2-576-amd64.deb
and did not receive any errors during installation. However I received the same error as above with verify-installation. I then tried doing conda update --all -y. This also did not fix the problem.
Here is some useful information:
(base) balter#rstudio:~$ which R
/home/balter/conda/bin/R
(base) balter#rstudio:~$ head -n3 /etc/rstudio/rserver.conf
# Server Configuration File
rsession-which-r=/home/balter/conda/bin/R
(base) balter#rstudio:~$ find . -name "libR.so"
./conda/lib/R/lib/libR.so
./conda/pkgs/r-base-4.1.3-h7880091_3/lib/R/lib/libR.so
(base) balter#rstudio:~$ sudo rstudio-server verify-installation
TTY detected. Printing informational message about logging configuration. Logging configuration loaded from '/etc/rstudio/logging.conf'. Logging to '/var/log/rstudio/rstudio-server/rserver.log'.
/usr/lib/rstudio-server/bin/rsession: error while loading shared libraries: libR.so: cannot open shared object file: No such file or directory

How to pip install different subdirectories from same repo?

I have an installation of Airflow in my requirements.in file -e git+git#github.com:apache/airflow.git#egg=apache-airflow and I want another line that will install the backport providers Google package with a cherry picked commit from my own fork.
-e git+git#github.com:myfork/airflow.git#egg=apache-airflow&subdirectory=airflow/operators/google but when compiling the requirements (piptools) I get an error saying that apache-airflow/airflow/providers/google cannot be found.
How can you pip install from the same repo under different directories? e.g. one for the full Airflow installation and the second for just the backport package under airflow.providers.google

Trying to install Tensorflow-Probability using reticulate - how to specify user?

In order to install the greta package, I need to first install tensorflow-probability. Tensorflow and Python 3.7 Anaconda are already installed.
When I attempt the following command in R: reticulate::conda_install("r-tensorflow", "tensorflow-probability", pip = TRUE), I get the following:
Could not install packages due to an EnvironmentError: [WinError 5] Access is denied:
'C:\\Users\\PHILTE~1\\ANACON~1\\envs\\r-tensorflow\\Lib\\site-packages\\numpy\\core\\multiarray.cp36-win_amd64.pyd'
Consider using the `--user` option or check the permissions.
I have admin permissions on my laptop, so I'm unsure why I'm getting an access denied. But how do I add the --user command within the reticulate command?
This webpage fixed it all for me: http://preposterior.updog.co/november-8-2018-getting-your-computing-environment-ready-for-greta.html#november-8-2018-getting-your-computing-environment-ready-for-greta
Steps:
Install Anaconda
Open Anaconda Prompt and run the following:
conda activate r-tensorflow
pip install --no-dependencies tensorflow-probability==0.4.0
conda install tensorflow=1.11
pip install h5py pyyaml requests Pillow scipy
Back in R, install the greta package from CRAN.
From version 0.4.0, greta now includes a install_greta_deps() function that installs all the relevant python modules required for greta.
You can learn more here:
https://greta-stats.org/articles/get_started.html

Install Emacs without root privilege under my personal directory, error while running ./configure --prefix -my-personal-directory

I am using system RedHat 6.8 and I tried to install emacs under my personal directory as I don't have permission to run root command.
Below is my command:
cd emacs-24.5
./configure --prefix -my-home-directory
Then I got error message as below:
configure: error: The following required libraries were not found:
libXpm libgif/libungif libtiff
Maybe some development libraries/packages are missing?
If you don't want to link with them give
--with-xpm=no --with-gif=no --with-tiff=no
as options to configure
I did some google search and it told me I need to install dependence libXpm, libgif/libungif and libtiff, but all instructions online were teaching you how to install these dependence with root privilege, which I don't have.
Then I downloaded source code of libXpm from https://xorg.freedesktop.org/archive/individual/lib/ and run following command
tar xvf libXpm-3.5.12.tar.gz
cd libXpm-3.5.12
configure --prefix -my-home-directory
make
make install
Then the libXpm successfully installed and I also appended the path to the libXpm in $LD_LIBRARY_PATH by
export LD_LIBRARY_PATH=/path/to/libXpm:$LD_LIBRARY_PATH
However, after that I tried to ./configure emacs again it still output that error
configure: error: The following required libraries were not found:
libXpm libgif/libungif libtiff
Maybe some development libraries/packages are missing?
If you don't want to link with them give
--with-xpm=no --with-gif=no --with-tiff=no
as options to configure
Anybody can help me out?

GLPK: No such file or directory error when trying to install R package

I am trying to install sparkTable in R 3.1.0 which depends on Rglpk. I manually installed GPLK on the system and added the libs folder to LD_LIBRARY_PATH before going into R for the install.packages("sparkTable") procedure. I get this error during the installation process. Any ideas?
* installing *source* package ‘Rglpk’ ...
** package ‘Rglpk’ successfully unpacked and MD5 sums checked
** libs
/bin/sh: line 0: cd: GLPK: No such file or directory
make: *** [GLPK.ts] Error 1
ERROR: compilation failed for package ‘Rglpk’
* removing ‘/opt/R/R-3.1.0/lib64/R/library/Rglpk’
ERROR: dependency ‘Rglpk’ is not available for package ‘sparkTable’
* removing ‘/opt/R/R-3.1.0/lib64/R/library/sparkTable’
sudo apt-get install libglpk-dev
did the trick for me.
I had this problem and took a good bit of digging in the package to understand what was happening. If Rgplk can't compile its test program when installing, it does something weird, including this bizarre cd to nowhere. Assuming glpk-devel is installed, the reason it can't compile the test program is that it can't find the gplk header as it is in a non-standard directory.
Just set the environment variable CPATH=/usr/include/glpk
and the test program will compile, allowing the package install to proceed normally.
I had this problem too. The following steps solved this issue for me. My current setup:
OS: Scientifc Linux version 6.5 (on a High Performance Cluster Server)
local user, no root access.
GLPK was not installed
Install GLPK in a local directory:
wget http://ftp.gnu.org/gnu/glpk/glpk-4.54.tar.gz
tar xfzv glpk-4.54.tar.gz
mkdir GLPK
cd glpk-4.54
./configure --prefix=/home/<username>/GLPK
make
make install
Install Rglpk (0.6-3):
cd ~
wget http://cran.r-project.org/src/contrib/Rglpk_0.6-3.tar.gz
export LIBRARY_PATH=/home/<username>/GLPK/lib
R CMD INSTALL Rglpk_0.6-3.tar.gz
I landed on this page, because I could not update igraph under Windows 10/11, since igraph also requires glpk as per OP. Specifically:
igraph_glpk_support.h:36:10: fatal error: glpk.h: No such file or
directory #include <glpk.h>
^~~~~~~~ compilation terminated. make: *** [C:/PROGRA~1/R/R-4.1.1/etc/x64/Makeconf:238: feedback_arc_set.o] Error
1 ERROR: compilation failed for package 'igraph'
removing 'C:/Users/xxx/Documents/R/win-library/4.1/igraph'
restoring previous 'C:/Users/xxx/Documents/R/win-library/4.1/igraph' Warning
in install.packages : installation of package ‘igraph’ had non-zero
exit status
I am under Windows 11 (but would be same for Windows 10).
Simple resolution is suggested here:
start Rtools Bash (found in all apps, Rtools 4.0 in the Windows menu)
Run pacman -S mingw-w64-x86_64-glpk and confirm with yes (y)
Run pacman -S mingw-w64-x86_64-libxml2 and confirm with yes (y)
Updating igraph in Rstudio now leads to a clean
DONE (igraph)
(there is no need to install anything, add any path, etc... just the above 4 steps)
In ubuntu 14.04, all above doesn't work. the following however works, without the need of installing libglpk-dev using apt-get.
download the glpk package from gnu and extract it:
wget http://ftp.gnu.org/gnu/glpk/glpk-4.55.tar.gz
tar xvf glpk-4.55.tar.gz
make a GLPK directory in your local path:
mkdir ~/GLPK
configure within glpk:
cd glpk-4.55
./configure --prefix=$HOME/GLPK
cd ..
export LD_LIBRARY_PATH=$HOME/GLPK/lib
export LIBRARY_PATH=$HOME/GLPK/lib
export CPATH=$HOME/GLPK/include
download the Rglpk package from cran and extract it:
wget http://cran.r-project.org/src/contrib/Rglpk_0.6-0.tar.gz
tar xvf Rglpk_0.6_0.tar.gz
move the glpk directory into Rglpk/src and rename it to GLPK:
mv glpk-4.55 Rglpk/src/GLPK
now you can install:
R CMD INSTALL Rglpk
now a bit of explanation of what's going on. The "src/Makevars.in" file in the Rglpk package contains a line of code to enter a non-existing directory 'GLPK' within the src/ folder:
(line 11 of Makevars.in)
GLPK.ts:
#(cd GLPK && make)
touch $#
this is where the problem arises. obviously the code is trying to build glpk within that directory for some unknown reasons. and the solution above is achieved simply by moving the downloaded (and configured) glpk directory there...
If you're on a Debian-based Linux distribution, run this to install GLTK:
sudo apt install libglpk-dev
If you're on MacOS, run this to install GLTK:
brew install gltk
Finally, install the R library:
install.packages("Rglpk")
On Centos, have sudo rights. None of the above worked, but had to install GLPK in /usr/local as suggested in this SO answer. Been trying to install it for about 3 hours fml
You must install glpk dependency first.
On macOS (via homebrew):
brew install glpk
or in RStudio (via homebrew):
system("brew install glpk")
It worked for me when I combined the answers from Simón Ramírez Amaya and shadowleaves:
wget http://ftp.gnu.org/gnu/glpk/glpk-4.54.tar.gz
tar xfzv glpk-4.54.tar.gz
mkdir GLPK
cd glpk-4.54
./configure --prefix=$HOME/GLPK
make
make install
cd ..
export LD_LIBRARY_PATH=$HOME/GLPK/lib
export LIBRARY_PATH=$HOME/GLPK/lib
export CPATH=$HOME/GLPK/include
wget https://cran.r-project.org/src/contrib/Rglpk_0.6-4.tar.gz
R CMD INSTALL Rglpk_0.6-4.tar.gz
To load the library
dyn.load(file.path(Sys.getenv("HOME"), "GLPK", "lib", "libglpk.so"))
library(Rglpk)
The answer of Zhiying Cui is right, however, if your script has some special codes, such as foreach for parallel computing, it may raise an error.
A better way is as fellow
If your os is Centos, try
yum install glpk-devel
then just go to R or Rstudio server and
install.packages("Rglpk")

Resources