I am having quite a bit of trouble setting up Theano to work with my graphics card - I hope you guys can give me a hand.
I have used CUDA before and it is properly installed as would be necessary to run Nvidia Nsight. However, I now want to use it with PyDev and am having several problems following the 'Using the GPU' part of the tutorial at http://deeplearning.net/software/theano/install.html#gpu-linux
The first is quite basic, and that is how to set up the environment variables. It says I should 'Define a $CUDA_ROOT environment variable'. Several sources have said to create a new '.pam_environment' file in my home directory. I have done this and written the following:
CUDA_ROOT = /usr/local/cuda-5.5/bin
LD_LIBRARY_PATH = /usr/local/cuda-5.5/lib64/lib
I am not sure if this is exactly the way it has to be written - apologies if this is a basic question. If I could get confirmation that this is indeed the correct place to have written it, too, that would be helpful.
The second problem is in the following part of the tutorial. It says to 'change the device option to name the GPU device in your computer'. Apparently this has something to do with THEANO_FLAGS and .theanorc, but nowhere am I able to find out what these are: are they files? If so where do I find them? The tutorial seems to be assuming some knowledge that I don't have!
Thanks for taking the time to read this: any and all answers are greatly appreciated - I am very much completely stuck at the moment!
On Linux/OSX:
Edit or create the file ~/.theanorc. The file should contain:
[global]
floatX = float32
device = gpu0
[nvcc]
fastmath = True
[cuda]
root=/usr/local/cuda-5.5/
# On a mac, this will probably be /Developer/NVIDIA/CUDA-5.5/
You need to add cuda to the $LD_LIBRARY_PATH variable. If you're running eclipse, you can go to Project properties > Interpreters > Configure and interpreter ... > Environment, and then add an LD_LIBRARY_PATH variable that points to your cuda lib folder (probably /Developer/NVIDIA/CUDA-5.5/lib64)
Now when you import theano it should print a message about finding the gpu. You can run the test code at http://deeplearning.net/software/theano/tutorial/using_gpu.html to see if it's using the gpu.
THEANO_FLAGS is an environment variable and .theanorc is a configuration file. You can use both mechanism to configure Theano. This is described here.
I never heard of the .pam_environment file. Also, you shouldn't just override the value of LD_LIBRARY_PATH, but append/prepend to it like this:
LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/usr/local/cuda-5.5/lib64/lib
For Theano, if you define CUDA_ROOT, you don't need to modify LD_LIBRARY_PATH, so I would just remove the last line.
Normally, if your shell is bash, people define the env variable CUDA_ROOT in the .bashrc file like this:
export CUDA_ROOT=/usr/local/cuda-5.5/bin
The change to .bashrc will only be used if you log out and log it again.
Related
I've recently switched to using PlatformIO for developing for STM32 using the following workflow:
Create a .ioc file using the CubeMX utility
Generate source code and the PlatformIO configuration from that .ioc file from the stm32pio command line utility
Edit, build, and debug using the PlatformIO plug-in for VSCode (Mac)
Overall, this works very well. However, I was previously using the CubeMX code generation in ST's CubeMX IDE, which placed a .s file in the source directory that (as I understand it) defined the NVIC, as well as the default function that was used for exceptions/interrupts that are not explicitly defined (i.e., those handled by their default weak implementations.) I don't see where this is defined in the new workflow. Is this generated dynamically as part of the build process?
The reason I'm asking is (beside wanting a better understanding of the process overall), I'd like to write an interrupt handler for EXTI0, but trigger it as a software interrupt, and not assign a pin to it. If that is not possible, then perhaps the entire point is moot.
I was able to find the answer. These steps might be useful to somebody else who comes across this question. This was done on MacOS, but should be similar to the process for other operating systems.
During the build process, the filename can be seen. It will be prefaced with startup_, followed by the name of the particular chip you're developing for. In my case, the line is
Compiling .pio/build/disco_f072rb/FrameworkCMSISDevice/gcc/startup_stm32f072xb.o
Searching in the .platformio folder of my user directory, I found the matching .s file, which in my case was .platformio/packages/framework-stm32cube/f0/Drivers/CMSIS/Device/ST/STM32F0xx/Source/Templates/gcc/startup_stm32f072xb.s
The structure of the path leading to the file indicates the particulars of the hardware and frameworks I'm using: STM32Cube framework, a F0 series chip, and the GCC compiler. The easiest way to find this file, and how I was able to find it, is using the find command from the terminal to search the PlatformIO directory.
Reading this file gives the lines I was looking for, defining the names of the functions to be used for exception and interrupt handling, such as the following:
.weak EXTI0_1_IRQHandler
.thumb_set EXTI0_1_IRQHandler,Default_Handler
It seems like, while I am using the CubeMX HAL for some drivers, the basic startup code is done using CMSIS, so it should be the same for HAL, LL, or CMSIS based builds.
I have an SCons project (an implementation of the Generic Mapping Tools tutorial at http://gmt.soest.hawaii.edu/doc/latest/GMT_Tutorial.html using SCons rather than shell scripts), and I am using Eclipse Neon to edit the Sconstruct file.
The Sconstruct file starts in quite a standard way (the rest of the file is immaterial to this question).
import os
import collections
env = Environment(ENV = os.environ)
bld = Builder(action = 'ps2pdf $SOURCE $TARGET', \
suffix = '.pdf', \
src_suffix = '.ps')
What is annoying me is that while the build works perfectly using scons, Eclipse keeps marking the Environment and Builder constructions as "Undefined variables".
I installed the SConsolidator plugin, but it makes no difference.
I find the marking of an error that is not an error incredibly annoying.
While I could do something like tell Eclipse to ignore the error, I would prefer something more intelligent, such as adding Scons to the library path. I have tried adding C:\Python27\Lib\site-packages\scons-2.5.1\Scons and C:\Python27\Lib\site-packages\scons-2.5.1\Scons\Scripts to the Python Interpreter Paths (Window → Preferences → PyDev → Interpreters → Python Interpreters → Paths), and using an import directive like from SConscript.SCons import * but it doesn't make a difference.
Try library path:
C:\Python27\Lib\site-packages\scons-2.5.1\
Then
from SCons.Script import *
First thing first - if you know exactly where does your plugin keep it's symbol index(es) make sure that they don't get deleted by whatever the editor/IDE/build think they are doing to "be helpful".
Python plugins for editors easily get confused and you may need you to draw very explicit and detailed picture and train them a bit (by stopping the code in debugger) in order for a scanner/indexer to "wake up" and finally scan and index all symbols.
What I do in VS (with PTVS) is to make one proj for Scons_lib (home at C:\Python27\Lib\site-packages\scons-2.5.0), another for Scons_Scripts (C:\Python27\Scripts\, startup file: scons.py)
and then separate projects for separate scons driven folders/builds. To get editor to immediately recognize building files as Python I add extension .scons (google's convention used in Chrome build, Sconcsript's are ProjName.scons), tell VS that .scons is also Python, rename Sconstruct to Make.scons.
Then I run in debugger, (with -f Make.scons -n of course) as many times as I can :-) trying to place breakpoints in different files form different sub-packages that I know will run. Letting scons throw exceptions for nothing (like missing Sconstuct file) is also file - the goal is to force indexer to go places because it sees that debugger is going there.
After N runs, (and/or K days). PTVS (you can consider it a plugin) all of the sudden starts recognizing all symbols, packages, sub-packages, only can't penetrate things explicitly hidden behind caches.
Gave up trying to determine which events exactly make the scanner/indexer see the light. Probably something like seeing the number of files at "unexpected" paths being in debugger. Most symbols are visible in debugger immediately, except for the files that load first (scons.py and my own file that I exec from scons.py to inject whatever I want before anything else runs.)
Also I keep PYTHONDONTWRITEBYTECODE=1 to make sure that it always has to load actual .py-s
It might potentially help to have a single file that would exercise something form every sub-package, litter it with breakpoints and let it be run a few times.
I have a problem with programming in Haxe. Whenever I try to run a code, I get an error:
Please don't add haxe/std to your classpath, instead set HAXE_STD_PATH env var.
I tried to find a solution - according to multiple sites, I'm supposed to change the environment variable HAXE_LIBRARY_PATH to HAXE_STD_PATH and set the value to actual path to the std file in my haxe folder. In my case, there was no such variable among my env vars, so I made HAXE_STD_PATH as a new one - and I still get the same error.
I'm using FlashDevelop and when I try to search for any class that would reference std, I can see in Type Explorer that there is a classpath that leads straight to haxe/std, which should not happen IMO. I looked at classpaths in Properties and the path to std was not listed, so I could not change it.
So problem could be in the classpath or in FlashDevelop settings.
Did anyone have the same problem?
Oh, and BTW I'm using Windows - and yes, restarting did not help.
I removed all: FD including program settings and HaxeToolkit (the whole map).
Then I reinstalled FD, Haxe (using the apps window from FD), Lime and openFL (from command window).
Now I don't get this error anymore.
there's definitely something wrong with setting up my environment variables.
I had messed around with it when I was trying to do an android app.
now i'm trying to set up netbeans, but it won't build.
when I first tried to build, tool selection came out, and it had looked for cygwin bin files in D: instead of C:, so I manually browsed for each file.
i found g++ make etc. however, I couldn't find gfortran.exe.
now, a message says that it can't find shell, and asks me to install cygwin.
this is what i have in my path:
D:\DAC driver and stuff\;%JAVA_HOME%\bin;
D:\matlab\MATLAB\runtime\win64;
D:\matlab\MATLAB\bin;C:\cygwin64\bin
and yes, cygwin is in C:\cygwin64
this is a picture of everything i have in my environment variables:
http://i567.photobucket.com/albums/ss114/samio_130/cake/pictwoo.png
edit: i think i deleted something from 'path'! should there be something infront of what i currently have??
%JAVA_HOME%\bin is no legit. Everything after it doesn't count, fix is in link.
http://forums.netbeans.org/viewtopic.php?p=151733#151733
I want to read data from GtkListStore and build an excel by phpExcel. First, I build a php file according to the GtkListStore, then I use php and phpexcel lib to compile and build execl file.
In my gtk code (compile in MinGW environment), I use execvp(cmd[0], (const char **)cmd); to call the external command -- php. In fact, cmd[0] is php.exe and cmd[1], cmd[2] ... are the parameters for php. After calling the php command, my gtk main window is closed and it quits my gtk program.
How can I prevent the php command from quitting main program? Should I use something else instead of execvp? Thank you.
execvp() and friends replace the current process with another process, so it's no surprise that your program quits. Use g_spawn_sync() or a related function - that will run your php program, then return control to your original program.
Let me answer my own question. Followed is the summary to what I have googled and tried within the past few days. It has nothing new but maybe is useful to a newbie like me.
First, thank ptomato. To avoid my problem in GTK, it must use g_spawn_sync or related functions. If your command is absolute path, you don't have to use the flag G_SPAWN_SEARCH_PATH, otherwise, make sure to use the flag.
Following are something related in MinGW environment.
-> gspawn-win32-helper.exe
In MinGW, to use g_spawn_sync, it must have gspawn-win32-helper.exe installed. When I installed GTK environment, I only extracted the useful lib or exe file I think it is useful then I missed gspawn-win32-helper.exe and it resulted in the problem -- Failed to execute helper program (No such file or directory) mentioned in the above comments. After extracting gspawn-win32-helper.exe from ftp://ftp.gtk.org/pub/glib/2.10/win32/glib-2.10.0.zip and installing it, g_spawn_sync worked.
-> For canonical Windows paths, both double backslashes and single forward slash work, e.g., c:/foo/bar and c:\foo\bar work.