I've spent an entire day on StackOverflow and the Tensorflow Github issues page trying to fix this, but still can't seem to resolve the problem, suggesting there's something obvious I'm missing.
I'm trying to get Tensorflow with GPU support running within RStudio on Ubuntu LTS 20.04. I've followed the RStudio installation instructions with Nvidia drivers 470, CUDA 11.2, and cuDDN 8.1.0, updated ~/.bashrc to point at
export CUDA_HOME=/usr/local/cuda
export PATH="/usr/local/cuda/bin:$PATH"
export LD_LIBRARY_PATH=/usr/lib/cuda/lib64:$LD_LIBRARY_PATH
export LD_LIBRARY_PATH=/usr/lib/cuda/include:$LD_LIBRARY_PATH
and installed TF with GPU support via install_keras(tensorflow = "gpu"), yet still getting the dreaded error:
W tensorflow/stream_executor/platform/default/dso_loader.cc:64] Could not load dynamic library 'libcudart.so.11.0'; dlerror: libcudart.so.11.0: cannot open shared object file: No such file or directory; LD_LIBRARY_PATH: /home/nickopotamus/.local/share/r-miniconda/envs/r-reticulate/lib:/usr/lib/R/lib::/lib:/usr/lib/x86_64-linux-gnu:/usr/lib/jvm/default-java/lib/server
sudo find / -name 'libcudart.so.11.0' finds the file in:
/home/nickopotamus/.local/share/r-miniconda/envs/r-reticulate/lib/libcudart.so.11.0
/home/nickopotamus/anaconda3/pkgs/cudatoolkit-11.3.1-h2bc3f7f_2/lib/libcudart.so.11.0
/home/nickopotamus/anaconda3/pkgs/cudatoolkit-11.2.0-h73cb219_8/lib/libcudart.so.11.0
/home/nickopotamus/anaconda3/pkgs/cudatoolkit-11.2.72-h2bc3f7f_0/lib/libcudart.so.11.0
/usr/local/cuda-11.2/targets/x86_64-linux/lib/libcudart.so.11.0
The top entry at least appears to be in the path that the error is searching, so I'm at a bit of a loss as to what to try next. Is it a conflict with the other anaconda packages (which I can't seem to remove), or am I simply being oblivious to a quick-fix?
Edit: Solved by using the NVIDIA Data Science Stack to install everything from scratch.
Not a particularly satisfying answer, but solved by removing all NVIDIA packages and installing the NVIDIA Data Science Stack from scratch, then using its Conda environment to run keras in R.
Related
Does anyone know how to use tensorflow in RStudio Cloud without running into this known fatal error? Are there versions of Python, Miniconda, TensorFlow, and Keras that make it work? My current setup script is here.
I use an RStudio Cloud project to teach a workshop. I am running into the known bug described here and here, and it is only a matter of time before the amazing folks at RStudio release a new IDE with a patch. But in the meantime, I have several workshops to teach, and I cannot install a different version of the IDE on Cloud. Cloud would be perfect if I can get it to work.
I already posted here about this, but I am trying to cast a wider net to see if anyone found a clever workaround.
I figured it out! All I needed was to downgrade to TensorFlow 1.13.1. I thought I already tried that, but then I realized install_keras() automatically upgrades TensorFlow unless I supply a version string to the tensorflow argument. Here is how I set up a local Python environment with TensorFlow 1.13.1 in RStudio Cloud.
reticulate::install_miniconda("miniconda")
Sys.setenv(WORKON_HOME = "virtualenvs")
reticulate::virtualenv_create("r-reticulate", python = "miniconda/bin/python")
keras::install_keras(
method = "virtualenv",
conda = "miniconda/bin/conda",
envname = "r-reticulate",
tensorflow = "1.13.1",
restart_session = FALSE
)
# Now add WORKON_HOME=/cloud/project/virtualenvs to .Renviron
First of all, I am working on a Mac. I am trying to install Blotter from GitHub. I found several descriptions of how to do that but my RStudio tells me that I am missing Building tools and gives me a link (https://www.cnet.com/how-to/install-command-line-developer-tools-in-os-x/) where it is described to do that. So far so good. I downloaded Xcode and the command line tools for Mac and installed those. Nothing changed even after restarting R. Then I found this https://cran.r-project.org/bin/macosx/tools/. I installed it and during that, it told me that I had to do the following
"This package will install clang 6.0.0 for OS X 10.11 (El Capitan) or higher with OpenMP support in /usr/local/clang6
In order to use this compiler you have to add /usr/local/clang6/bin to the PATH environment variable such as
export PATH=/usr/local/clang6/bin:$PATH"
So I changed the environmental variable path as follows http://blog.tonytsai.name/blog/2018-05-07-setting-path-variable-for-gs-command-in-rstudio/.
How I changed the PATH variable.
Again I restarted R but still, nothing changed. I still get the notice that the building tool is missing.
Somehow it seems to me that I installed everything correctly but R doesn't recognize the Programmes. Does anyone have an idea? I tried to search for settings to tell R that I installed the command line tool but couldn't really find anything helpful.
Ok, a bit of an update.
Best I can see it that Blotter is built and stored on R-Forge packages under a package called RStrategist
In R console type/cut & paste this.
install.packages('RStrategist',repos='http://R-Forge.r-project.org')
See R forR-Forge for more details. Once this has been installed run instead.
library(RStrategist)
Unfortunately, I am not willing to install this package and see if it works mainly because 1) don't need it nor know how to use it, 2) not sure how good packages are from R-forge, though it seems legit, but, this brings me back to point one.
So before i read the updated answer of Conrad Thiele i was trying around bit. Basically i deleted R, R Studio, Xcode and Command Line tools. Then i installed Xcode, Command Line tools, R and RStudio. Then i followed the notice on https://cran.r-project.org about the tools and installed both stated tools. As mentioned in the original question the Clang package tells you to change the Environmental Variable. And there was the mistake i believe. I originally simply pasted "PATH=/usr/local/clang6/bin:$PATH" into the the ".Renviron" file. With reading up online i noticed that "export PATH=/usr/local/clang6/bin:$PATH" is actually a Command for the Mac Terminal. After executing it, it sill didn't work but then i remembered that i still had the Path "PATH=/usr/local/clang6/bin:$PATH" in the the ".Renviron" file. Once i deleted that it worked. So i guess the key was that with changing the Environmental Variable correctly R found the connection with the right tool. Patients paid off.
I just updated R from version 2.15.1 to version 3.0.0 on my MAC running 10.6.8 and now R crashes on startup.
I get the error:
Error in getLoadedDLLs() : there is no .Internal function 'getLoadedDLLs'
Error in checkConflicts(value) :
".isMethodsDispatchOn" is not a BUILTIN function
Any ideas on how to go about?
The most common cause of this is having a corrupted ".Rdata" file in your working directory. Using the Mac Finder.app you will not by default be able to see files that begin with a ".", so-called dotfiles. Those files can be "seen" if you execute a change to the plist controlling the behavior of Finder.app. Open a Terminal.app window and run this bit of code:
defaults write com.apple.Finder AppleShowAllFiles YES
Then /point/-/click/-/hold/ on Dock-Finder-icon, and choose "Relaunch"
If you to do so, you can then change it back with the obvious modfication to that procedure. I happen to like seeing the hidden files so that's the way I run my Mac all the time, but some people may feel it is too dangerous to expose the "hidden secrets" to their own bumbling.
Paul raises a good point: I run the following R function in the R console after updating:
update.packages(checkBuilt=TRUE, ask=FALSE)
I have a lot of installed packages and paging through the entire list has gotten too tiresome so I bypass the ask-messages. Sometimes you will get errors because there may be dependencies on r-forge or Omegahat packages or on packages that need to be compiled from source. These may need to be updated "by hand". And you may need more than one pass through such an effort. Take notes of which packages are missing and fill them in.
I had the same problem running RKWard on ubuntu 12.04.
Check your r-base-core, like Paul suggested, to make sure the version is also at the latest version. Mine didn't update automatically. I had a platform dependent version, but RKWard was calling the new version. To solve this problem, I simply marked r-base-core for removal and reinstalled the latest version or r-base-core. poof problem fixed, bippity boppity boo!
I suspect that your error is similar to mine because I had also JUST updated RKWard. Start at updating r-base-core or try to get all of the dependencies to match up the versions.
I hope that you can translate this into what to do on a MAC,
SU
In order to use a specific library that has not been updated for some time, I want to use an older version of R (2.3.1), under linux Mint 14.
I got the source file, installed the required library; checking with :
apt-cache showsrc r-base | grep Build-Depends
and issued, as indicated in the R-admin help page, the command:
./configue
that ended without error; then
make
that terminated with the following error message:
In file included from datetime.c:95:0:
Rstrptime.h:201:12: erreur: conflicting types for ‘wcsncasecmp’
In file included from ../../src/include/Defn.h:928:0,
from datetime.c:58:
/usr/include/wchar.h:172:12: note: previous declaration of ‘wcsncasecmp’ was here
Does anyone know what trigered that error (conflicting type between files datetime.c and wchar.h, if I understand well), and how I could keep compiling past this error.
Thanks in advance for your help.
The problem is that R 2.3.1 is very old, and was developed with the old C libraries in mind. With a recent linux install, you have the new C libraries which might not work well with your old R version. What you could do:
Install an old version of linux from around the time of the R version, for example in a virtual machine.
Port the package to the new version of R yourself.
The second option takes more time, but will make the work you base on the package more future proof.
Yesterday I removed R2.11 from my system (Win7, 64bit), since I´m working on R2.13.
Since then i get an error message:
> require(rJava)
Lade nötiges Paket: rJava
Error : .onLoad in loadNamespace() fehlgechlagen, Details:
Aufruf: rJava
Fehler: inDL(x, as.logical(local), as.logical(now), ...)
I tried specifying PATH, since I found on the internet that it might have something to do with jvm.dll:
c:\Rtools\bin;
c:\Rtools\perl\bin;
c:\Rtools\MinGW\bin;
c:\Rtools\MinGW64\bin;
C:\Windows\system32;
%R_HOME%\bin;
C:\Program Files\R\R-2.13.0\bin;
C:\Program Files\Java\jre6\bin\server
However I could not solve the problem...
I also can´t run R from the win command line (just type "R"?)
Any suggestions?
Here is some quick advice on how to get up and running with R + rJava on Windows 7 64bit. There are several possibilities, but most have fatal flaws. Here is what worked for me:
Add jvm.dll to your PATH
rJava, the R<->Java bridge, will need jvm.dll, but R will have trouble finding that DLL. It resides in a folder like
C:\Program Files\Java\jdk1.6.0_25\jre\bin\server
or
C:\Program Files\Java\jre6\jre\bin\client
Wherever yours is, add that directory to your windows PATH variable. (Windows -> "Path" -> "Edit environment variables to for your account" -> PATH -> edit the value.)
You may already have Java on your PATH. If so you should find the client/server directory in the same Java "home" dir as the one already on your PATH.
To be safe, make sure your architectures match.If you have Java in Program Files, it is 64-bit, so you ought to run R64. If you have Java in Program Files (x86), that's 32-bit, so you use plain 32-bit R.
Re-launch R from the Windows Menu
If R is running, quit.
From the Start Menu , Start R / RGUI, RStudio. This is very important, to make R pick up your PATH changes.
Install rJava 0.9.2.
Earlier versions do not work! Mirrors are not up-to-date, so go to the source at www.rforge.net: http://www.rforge.net/rJava/files/. Note the advice there
“Please use
`install.packages('rJava',,'http://www.rforge.net/')`
to install.”
That is almost correct. This actually works:
install.packages('rJava', .libPaths()[1], 'http://www.rforge.net/')
Watch the punctuation! The mysterious “.libPaths()[1],” just tells R to install the package in the primary library directory. For some reason, leaving the value blank doesn’t work, even though it should default.
I finally solved the problem:
It seems that rJava searches for jvm.dll in ~\Java\jre6\bin\client.
However this folder didn´t exist on my system (jvm.dll was in ~\bin\server).
So I just made a copy of jvm.dll in a folder ~\bin\client\ and added this to the path.
Now everything works fine!
My problem was solved by
install.packages("SqlRender",INSTALL_opts="--no-multiarch")
It was a package that depends on rJava and all advices were telling me to fix Java installation. But the solution was to use install option that simply forgets about i386 architecture. (also works with drat library and packages not from CRAN)
This may be due to a conflict between RStudio and Java versions. If you have installed 64 bit java and RStudio is running in 32 bit mode, you may experience problems like this. As a solution, you can change the 32-64 bit selection in the Tools-> Global Options-> General section in RStudio.
You can find detailed information here.
In my case installing proper version of Java solved my problem.
I installed 64x bit java, cause I use 64x bit R version.
I solved it by following these steps
setting my environment Sys.setenv(JAVA_HOME='C:\\Program Files (x86)\\Java\\jre6')
Manually installing rJava package from install package (even this should work:
install.packages('rJava', .libPaths()[1], 'http://www.rforge.net/'))
library(rJava)
I solved this problem as follows. I've been trying for 2 days.
Windows 7 users do not write ... \ bin \ x64 in environment variables.
Instead, define the path as follows.
JAVA_HOME "C: \ Program Files \ Java \ jre1.8.0_251"
R_HOME C: \ Program Files \ R \ R-3.5.3
In RStudio type .LibPaths()
This will give you an path in you windows system where your library’s are located
Go there and delete rJava. If it is being used by applications of Java, kill all Java programs in the Task Manager.
Go to computer and properties, click on change environment variables
Edit JAVA_HOME and all Java related paths to the path where your newest installation of Java is located and save.