SYSTEM: Ryzen 5800x, rx 6700xt, 32 gigs of RAM, Ubuntu 22.04.1
I'm attempting to install Stable-Diffusion by following https://youtu.be/d_CgaHyA_n4
When attempting to run the SD script, I get the "hipErrorNoBinaryForGpu: Unable to find code object for all current devices!" error.
I believe this is caused by PyTorch not working as expected. When validating Pytorchs' installation with "The Master Test", I get the same error:
"hipErrorNoBinaryForGpu: Unable to find code object for all current devices!"
Aborted (core dumped)
I believe that it is install correctly as using the conda list command tells me that torch 1.12.0a0+git2a932eb and torchvision 0.13.0a0+f5afae5 are installed. Interestingly, when I change the command ever-so-slightly to torch.cuda.is_available, (without the parentheses), I get the following output: <function is_available at 0x7f42278788b0>. Granted, I'm not sure what this is telling me. Following the "Verification" step resulted in the expected array of random numbers. However, failed the GPU driver check.
Thank you in advance.
Try running the following command:
export HSA_OVERRIDE_GFX_VERSION=10.3.0
This made it work on my machine using an RX 6600 XT, with which I got the same error running it, before exporting the variable.
I was struggling with ROCm enabled Tensorflow, the tensorflow-rocm, package. Without setting up the environment variable
export HSA_OVERRIDE_GFX_VERSION=10.3.0
, tensorflow-rocm crashes.
After setting it up, it works with 6700XT as well (at least it does not crash.) But still having problem with finding libraries.
It seems ROCm has a lot to catch up.
I'm trying to get a Lambda happy version of XPDF's pdftohtml to work but am having no luck.
So far the following has been tried:
Created Docker container running the latest amazonlinux image
I've copied the source code into this container and ran:
yum install cmake, gcc, gcc-c++, freetype-devel
Compiling the code with cmake produces a binary which executes perfectly in the container which should be the same OS and environment as Lambda.
I've verified the version of libc.so.6 as 2.26 within the container.
I've copied this into my AWS zip folder and included the following dependencies in a lib folder ready to upload:
libfreetype.so.6.10.0, libpng15.so.15, libstdc++.so.6.0.24
These dependencies are copied directly from the container used to compile the code.
Python function then connects these via
os.environ.update(dict(LD_LIBRARY_PATH='/var/task/lib'))
At the end of this, I run the function and get the following error code:
/var/task/pdftohtml: /lib64/libc.so.6: version `GLIBC_2.18' not found (required by /var/task/lib/libstdc++.so.6)
I've no idea where the GLIBC_2.18 comes from as this version isn't present in the container used to compile it.
Really stumped but keen to get it finished as it would produce a lightweight binary perfect for a Lambda function!
Where am I going wrong?
EDIT
SOLVED - see the comments below. There are two versions of AWS Linux and Lambda runs this version
I ran in an EC2 instance as one of the commenters suggested. Whilst the libstdc++.so.6.24 looked to be the right version, as it was itself compiled with a different GLIBC version, it throws an error. Compiling in EC2 from the source code worked fine. The other trick was making sure the CXX_FLAGS included -std=c++11. Thanks to those who contributed to help me solve this!
I've no idea where the GLIBC_2.18 comes from as this version isn't present in the container used to compile it.
I think you don't understand symbol version dependencies (see here).
The error message is telling you that your libstdc++.so.6 was built against GLIBC-2.18 or newer, and you are running against GLIBC-2.17 or older.
Where am I going wrong?
Your build environment is targeting something much newer than what your deployment environment contains.
You need to either find a built environment that matches your deployment target, or you need to change your deployment target to be not older than your build environment.
I wrote application for linux which uses Qt5.
But when I am trying to launch it on the linux without Qt SDK installed, the output in console is:
Failed to load platform plugin "xcb". Available platforms are:
How can I fix this? May be I need to copy some plugin file?
When I use ubuntu with Qt5 installed, but I rename Qt directory, the same problem occurs. So, it uses some file from Qt directory...
UPDATE:
when I create in the app dir "platforms" folder with the file libqxcb.so, the app still doesnot start, but the error message changes:
Failed to load platform plugin "xcb". Available platforms are:
xcb
How can this happen? How can platform plugin be available but can't be loaded?
Use ldd (man ldd) to show shared library dependencies. Running this on libqxcb.so
.../platforms$ ldd libqxcb.so
shows that xcb depends on libQt5DBus.so.5 in addition to libQt5Core.so.5 and libQt5Gui.so.5 (and many other system libs). Add libQt5DBus.so.5 to your collection of shared libs and you should be ready to move on.
As was posted earlier, you need to make sure you install the platform plugins when you deploy your application. Depending on how you want to deploy things, there are two methods to tell your application where the platform plugins (e.g. platforms/plugins/libqxcb.so) are at runtime which may work for you.
The first is to export the path to the directory through the QT_QPA_PLATFORM_PLUGIN_PATH variable.
QT_QPA_PLATFORM_PLUGIN_PATH=path/to/plugins ./my_qt_app
or
export QT_QPA_PLATFORM_PLUGIN_PATH=path/to/plugins
./my_qt_app
The other option, which I prefer is to create a qt.conf file in the same directory as your executable. The contents of which would be:
[Paths]
Plugins=/path/to/plugins
More information regarding this can be found here and at using qt.conf
I tried to start my binary, compiled with Qt 5.7, on Ubuntu 16.04 LTS where Qt 5.5 is preinstalled. It didn't work.
At first, I inspected the binary itself with ldd as was suggested here, and satisfied all "not found" dependencies. Then this notorious This application failed to start because it could not find or load the Qt platform plugin "xcb" error was thrown.
How to resolve this in Linux
Firstly you should create platforms directory where your binary is, because it is the place where Qt looks for XCB library. Copy libqxcb.so there. I wonder why authors of other answers didn't mention this.
Then you may want to run your binary with QT_DEBUG_PLUGINS=1 environment variable set to check which dependencies of libqxcb.so are not satisfied. (You may also use ldd for this as suggested in the accepted answer).
The command output may look like this:
me#xerus:/media/sf_Qt/Package$ LD_LIBRARY_PATH=. QT_DEBUG_PLUGINS=1 ./Binary
QFactoryLoader::QFactoryLoader() checking directory path "/media/sf_Qt/Package/platforms" ...
QFactoryLoader::QFactoryLoader() looking at "/media/sf_Qt/Package/platforms/libqxcb.so"
Found metadata in lib /media/sf_Qt/Package/platforms/libqxcb.so, metadata=
{
"IID": "org.qt-project.Qt.QPA.QPlatformIntegrationFactoryInterface.5.3",
"MetaData": {
"Keys": [
"xcb"
]
},
"className": "QXcbIntegrationPlugin",
"debug": false,
"version": 329472
}
Got keys from plugin meta data ("xcb")
loaded library "/media/sf_Qt/Package/platforms/libqxcb.so"
QLibraryPrivate::loadPlugin failed on "/media/sf_Qt/Package/platforms/libqxcb.so" : "Cannot load library /media/sf_Qt/Package/platforms/libqxcb.so: (/usr/lib/x86_64-linux-gnu/libQt5DBus.so.5: version `Qt_5' not found (required by ./libQt5XcbQpa.so.5))"
This application failed to start because it could not find or load the Qt platform plugin "xcb"
in "".
Available platform plugins are: xcb.
Reinstalling the application may fix this problem.
Aborted (core dumped)
Note the failing libQt5DBus.so.5 library. Copy it to your libraries path, in my case it was the same directory where my binary is (hence LD_LIBRARY_PATH=.). Repeat this process until all dependencies are satisfied.
P.S. thanks to the author of this answer for QT_DEBUG_PLUGINS=1.
I tried the main parts of each answer, to no avail. What finally fixed it for me was to export the following environment variables:
LD_LIBRARY_PATH=/usr/local/lib:~/Qt/5.9.1/gcc_64/lib
QT_QPA_PLATFORM_PLUGIN_PATH=~/Qt/5.9.1/gcc_64/plugins/
Ubuntu 16.04 64bit.
I got the problem for apparently no reasons. The night before I watched a movie on my VideoLan instance, that night I would like to watch another one with VideoLan. VLC just didn't want to run because of the error into the question.
I google a bit and I found the solution it solved my problem: from now on, VLC is runnable just like before. The solution is this comand:
sudo ln -sf /usr/lib/x86_64-linux-gnu/qt5/plugins/platforms/ /usr/bin/
I am not able to explain what are its consequencies, but I know it creates some missing symbolic link.
Since version 5, Qt uses a platform abstraction system (QPA) to abstract from the underlying platform.
The implementation for each platform is provided by plugins. For X11 it is the XCB plugin. See Qt for X11 requirements for more information about the dependencies.
There might be many causes to this problem. The key is to use
export QT_DEBUG_PLUGINS=1
before you run your Qt application. Then, inspect the output, which will point you to the direction of the error. In my case it was:
Cannot load library /opt/nao/plugins/platforms/libqxcb.so: (/opt/nao/bin/../lib/libz.so.1: version `ZLIB_1.2.9' not found (required by /usr/lib/x86_64-linux-gnu/libpng16.so.16))
But that is solved in different threads. See for instance https://stackoverflow.com/a/50097275/2408964.
Probably this information will help. I was on Ubuntu 18.04 and when I tried to install Krita, using the ppa method, I got this error:
This application failed to start because it could not find or load the Qt platform plugin "xcb" in "".
Available platform plugins are: linuxfb, minimal, minimalegl, offscreen, wayland-egl, wayland, xcb.
Reinstalling the application may fix this problem.
Aborted
I tried all the solutions that I found in this thread and other webs without any success.
Finally, I found a post where the author mention that is possible to activate the debugging tool of qt5 using this simple command:
export QT_DEBUG_PLUGINS=1
After adding this command I run again krita I got the same error, however this time I knew the cause of that error.
libxcb-xinerama.so.0: cannot open shared object file: No such file or directory.
This error prevents to the "xcb" to load properly. So, the solution will be install the `libxcb-xinerama.so.0" right? However, when I run the command:
sudo apt install libxcb-xinerama
The lib was already installed. Now what Teo? Well, then I used an old trick :) Yeah, that one --reinstall
sudo apt install --reinstall libxcb-xinerama
TLDR: This last command solved my problem.
I ran into a very similar problem with the same error message. First, debug some by turning on the Qt Debug printer with the command line command:
export QT_DEBUG_PLUGINS=1
and rerun the application. For me this revealed the following:
"Cannot load library /home/.../miniconda3/lib/python3.7/site-packages/PyQt5/Qt/plugins/platforms/libqxcb.so: (libxkbcommon-x11.so.0: cannot open shared object file: No such file or directory)"
"Cannot load library /home/.../miniconda3/lib/python3.7/site-packages/PyQt5/Qt/plugins/platforms/libqxcb.so: (libxkbcommon-x11.so.0: cannot open shared object file: No such file or directory)"
Indeed, I was missing libxkbcommon-x11.so.0 and libxkbcommon-x11.so.0. Next, check your architecture using dpkg from the linux command line. (For me, the command "arch" gave a different and unhelpful result)
dpkg --print-architecture #result for me: amd64
I then googled "libxkbcommon-x11.so.0 ubuntu 18.04 amd64", and likewise for libxkbcommon-x11.so.0, which yields those packages on packages.ubuntu.com. That told me, in retrospect unsurprisingly, I'm missing packages called libxkbcommon-x11-0 and libxkbcommon0, and that installing those packages will include the needed files, but the dev versions will not. Then the solution:
sudo apt-get update
sudo apt-get install libxkbcommon0
sudo apt-get install libxkbcommon-x11-0
So, I spent about a day trying to figure out what was the issue; tried all the proposed solutions, but none of that worked like installing xcb libs or exporting Qt plugins folder. The solution that suggested to use QT_DEBUG_PLUGINS=1 to debug the issue didn't provide me a direct insight like in the answer - instead I was getting something about unresolved symbols within Qt5Core.
That gave me a hint, though: what if it's trying to use different files from different Qt installations? On my machine I had standard version installed in /home/username/Qt/ and some local builds within my project that I compiled by myself (I have other custom built kits as well in other locations). Whenever I tried to use any of the kits (installed by Qt maintenance tool or built by myself), I would get an "xcb error".
The solution was simple: provide the Qt path through CMAKE_PREFIX_PATH and not though Qt5_DIR as I did, and it solved the problem. Example:
cmake .. -DCMAKE_PREFIX_PATH=/home/username/Qt/5.11.1/gcc_64
I faced the same problem when after installing Viber. It had all required qt libraries in /opt/viber/plugins/.
I checked dependencies of /opt/viber/plugins/platforms/libqxcb.so and found missing dependencies. They were libxcb-render.so.0, libxcb-image.so.0, libxcb-icccm.so.4, libxcb-xkb.so.1
So I resolved my issue by installing missing packages with this libraries:
apt-get install libxcb-xkb1 libxcb-icccm4 libxcb-image0 libxcb-render-util0
I like the solution with qt.conf.
Put qt.conf near to the executable with next lines:
[Paths]
Prefix = /path/to/qtbase
And it works like a charm :^)
For a working example:
[Paths]
Prefix = /home/user/SDKS/Qt/5.6.2/5.6/gcc_64/
The documentation on this is here: https://doc.qt.io/qt-5/qt-conf.html
All you need to do is
pip uninstall PyQt5
and
conda install pyqt
Most of the problem of pyqt can be fixed by this simplest solution.
In my case, I needed to deploy two Qt apps on an Ubuntu virtualbox guest. One was command-line ("app"), the other GUI_based ("app_GUI").
I used "ldd app" to find out what the required libs are, and copied them
to the Ubuntu guest.
While the command-line executable "app" worked ok, the GUI-based executable crashed, giving
the "Failed to load platform plugin "xcb" error. I checked ldd for libxcb.so, but this too had no missing dependencies.
The problem seemed to be that while I did copy all the right libraries I accidentally had copied also libraries that were already present at the guest system.. meaning that (a) they were unnecessary to copy them in the first place and (b) worse, copying them produced incompatibilities between the install libraries.
Worse still, they were undetectable by ldd like I said..
The solution? Make sure that you copy libraries shown as missing by ldd and absolutely no extra libraries.
In my case missing header files were the reason libxcb was not built by Qt. Installing them according to https://wiki.qt.io/Building_Qt_5_from_Git#Linux.2FX11 resolved the issue:
yum install libxcb libxcb-devel xcb-util xcb-util-devel mesa-libGL-devel libxkbcommon-devel
Folks trying to get this started on Ubuntu 20.04 please try to run this and see if this solves the problem. This worked for me
sudo apt-get update -y
sudo apt-get install -y libxcb-xinerama0
I link all Qt stuff statically to the generic Linux builds of my open source projects. It makes life a bit easier. You just need to build static versions of Qt libraries first. Of course this cannot be applied to closed source software due to licensing issues. The deployment of Qt5 apps on Linux is currently a bit problematic, because Ubuntu 12.04, for example, doesn't have Qt5 libraries in the package repositories.
I had this problem, and on a hunch I removed the Qt Configs from my environment. I.e.,
rm -rf ~/.config/Qt*
Then I started qtcreator and it reconfigured itself with the existing state of the machine. It no longer remembered where my projects were, but that just meant I had to browse to them "for the first time" again.
But more importantly it built itself a coherent set of library paths, so I could rebuild and run my project executables again without the xcb or qxcb libraries going missing.
I faced the same situation, but on a Ubuntu 20.04 VM.
TL;DR: Check file permissions.
What I did:
I copied the Qt libs required to /usr/local/lib/x86_64-linux-gnu/ and added it to LD_LIBRARY_PATH
I copied the platforms folder from Qt to my application directory and added it to QT_PLUGIN_PATH
I ran ldd on the executable and in the offending libqxcb.so (ldd libqxcb.so), and it complains about some dependencies although ldconfig listed them as found.
linux-vdso.so.1 (0x00007ffee19af000)
libQt5XcbQpa.so.5 => not found
libfontconfig.so.1 => /lib/x86_64-linux-gnu/libfontconfig.so.1 (0x00007f7cb18fb000)
libfreetype.so.6 => /lib/x86_64-linux-gnu/libfreetype.so.6 (0x00007f7cb183c000)
libz.so.1 => /lib/x86_64-linux-gnu/libz.so.1 (0x00007f7cb1820000)
libQt5Gui.so.5 => /usr/local/lib/x86_64-linux-gnu/libQt5Gui.so.5 (0x00007f7cb0fd4000)
libQt5DBus.so.5 => not found
I used export QT_DEBUG_PLUGINS=1 for further info. It complains about missing files, although they are there.
What I found:
For some reason, when copying to the VM through the shared folder the files permissions were not the correct ones.
Thus, I ran: sudo chmod 775 * on the libs and voilà.
I solved the issue through this https://github.com/NVlabs/instant-ngp/discussions/300
pip uninstall opencv-python
pip install opencv-python-headless
This seems to have been a problem with the cv2 Python package and how it loops in Qt
sudo ln -sf /usr/lib/...."adapt-it"..../qt5/plugins/platforms/ /usr/bin/
It creates the symbolic link it's missed. Good for QT ! Good for VLC !!