Differences

This shows you the differences between two versions of the page.

--- gpu [2018/04/27 18:22]
popel [Servers with GPU units]
+++ gpu [2018/06/21 10:04]
vodrazka [Servers with GPU units]
@@ Line 4: / Line 4: @@
 ===== Servers with GPU units =====
-GPU cluster ''gpu.q'' at Malá Strana:
+GPU cluster ''gpu-ms.q'' at Malá Strana:
 | machine | GPU type | GPU driver version | [[https://en.wikipedia.org/wiki/CUDA#GPUs_supported|cc]] | GPU cnt | GPU RAM (GB) | machine RAM (GB)|
-| dll1 |  GeForce GTX 1080 |  384.69 |  6.1 |  8 |  8 |  250 |
+| dll1 |  GeForce GTX 1080 |  396.24 |  6.1 |  8 |  8 |  249 |
-| dll2 |  GeForce GTX 1080 |  387.34 |  6.1 |  8 |  8 |  250 |
+| dll2 |  GeForce GTX 1080 |  396.24 |  6.1 |  8 |  8 |  249 |
-| dll3 |  GeForce GTX 1080 Ti |  375.66 |  6.1 |  9 |  11 |  250 |
+| dll3 |  GeForce GTX 1080 Ti |  396.24 |  6.1 |  9 |  11 |  249 |
-| dll4 |  GeForce GTX 1080 Ti |  375.66 |  6.1 |  10 |  11 |  250 |
+| dll4 |  GeForce GTX 1080 Ti |  396.24 |  6.1 |  10 |  11 |  249 |
-| dll5 |  GeForce GTX 1080 Ti |  384.69 |  6.1 |  10 |  11 |  250 |
+| dll5 |  GeForce GTX 1080 Ti |  396.24 |  6.1 |  10 |  11 |  249 |
-| dll6 |  GeForce GTX 1080 Ti |  384.69 |  6.1 |  9 |  11 |  122 |
+| dll6 |  GeForce GTX 1080 Ti |  396.24 |  6.1 |  9 |  11 |  123 |
-| titan-gpu |  GeForce GTX TITAN Z |  381.22 |  3.5 |  2 |  6 |  31 |
+| kronos |  GeForce GTX 1080 Ti |  396.24 |  6.1 |  1 |  11 |  123 |
-| iridium |  Quadro K2000 |  367.48 |  3.0 |  1 |  2 |  504 |
+| titan1 |  GeForce GTX 1080 |  396.24 |  6.1 |  1 |  8 |  30 |
-| kronos |  GeForce GTX 1080 Ti |  384.81 |  6.1 |  1 |  11 |  125 |
+| titan2 |  Tesla K40c |  396.24 |  3.5 |  1 |  11 |  30 |
-| titan |  GeForce GTX 1080 |  381.22 |  6.1 |  1 |  8 |  31 |
+| twister1 |  Tesla K40c |  396.24 |  3.5 |  1 |  11 |  45 |
-| twister1 |  Tesla K40c |  367.48 |  3.5 |  1 |  11 |  47 |
+| twister2 |  Tesla K40c |  396.24 |  3.5 |  1 |  11 |  45 |
-| twister2 |  Quadro P5000 |  367.48 |  6.1 |  1 |  17 |  47 |
 Desktop machines:
@@ Line 25: / Line 24: @@
 Not used at the moment: GeForce GTX 570 (from twister2)
-All machines have CUDA8.0 and should support both Theano and TensorFlow.
+Multiple versions of CUDA library are accessible on each machine together with cudnn. Theano and TensorFlow is supported.
+TODO - update link:
 [[https://ufaladm2.ufal.hide.ms.mff.cuni.cz/munin/ufal.hide.ms.mff.cuni.cz/lrc-headnode.ufal.hide.ms.mff.cuni.cz/index.html#dll|GPU usage rolling graphs]]
@@ Line 33: / Line 33: @@
   * First, read [[internal:Linux network]] and [[:Grid]].
   * All the rules from [[:Grid]] apply, even more strictly than for CPU because there are too many GPU users and not as many GPUs available. So as a reminder: always use GPUs via ''qsub'' (or ''qrsh''), never via ''ssh''. You can ssh to any machine e.g. to run ''nvidia-smi'' or ''htop'', but not to start computing on GPU. Don't forget to specify you RAM requirements with e.g. ''-l mem_free=8G,act_mem_free=8G,h_vmem=12G''.
-  * Always specify the number of GPU cards (e.g. ''gpu=1''), the minimal Cuda capability you need (e.g. ''gpu_cc_min3.5=1'') and your GPU memory requirements (e.g. ''gpu_ram=2G''). Thus e.g. <code>qsub -q gpu.q -l gpu=1,gpu_cc_min3.5=1,gpu_ram=2G</code>
+  * Always specify the number of GPU cards (e.g. ''gpu=1''), the minimal Cuda capability you need (e.g. ''gpu_cc_min3.5=1'') and your GPU memory requirements (e.g. ''gpu_ram=2G''). Thus e.g. <code>qsub -q gpu-ms.q -l gpu=1,gpu_cc_min3.5=1,gpu_ram=2G</code>
-  * If you need more than one GPU card (on a single machine), always require as many CPU cores (''-pe smp X'') as many GPU cards you need. E.g. <code>qsub -q gpu.q -l gpu=4,gpu_cc_min3.5=1,gpu_ram=7G -pe smp 4</code> **Warning**: currently, this does not work, so you can omit the ''-pe smp X'' part. Milan Fučík is working on a fix.
+  * If you need more than one GPU card (on a single machine), always require as many CPU cores (''-pe smp X'') as many GPU cards you need. E.g. <code>qsub -q gpu-ms.q -l gpu=4,gpu_cc_min3.5=1,gpu_ram=7G -pe smp 4</code> **Warning**: currently, this does not work, so you can omit the ''-pe smp X'' part. Milan Fučík is working on a fix.
-  * For interactive jobs, you can use ''qrsh'', but make sure to end your job as soon as you don't need the GPU (so don't use qrsh for long training). **Warning: ''-pty yes bash'' is necessary**, otherwise the variable ''$CUDA_VISIBLE_DEVICES'' will not be set correctly. E.g. <code>qrsh -q gpu.q -l gpu=1,gpu_ram=2G -pty yes bash</code>In general: don't reserve a GPU (as described above) without actually using it for longer time. (E.g. try separating steps which need GPU and steps which do not and execute those separately on our GPU resp. CPU cluster.) Ondřej Bojar has a script /home/bojar/tools/servers/watch_gpus for watching reserved but unused GPU on most machines which will e-mail you, but don't rely on in only.
+  * For interactive jobs, you can use ''qrsh'', but make sure to end your job as soon as you don't need the GPU (so don't use qrsh for long training). **Warning: ''-pty yes bash'' is necessary**, otherwise the variable ''$CUDA_VISIBLE_DEVICES'' will not be set correctly. E.g. <code>qrsh -q gpu-ms.q -l gpu=1,gpu_ram=2G -pty yes bash</code>In general: don't reserve a GPU (as described above) without actually using it for longer time. (E.g. try separating steps which need GPU and steps which do not and execute those separately on our GPU resp. CPU cluster.) Ondřej Bojar has a script /home/bojar/tools/servers/watch_gpus for watching reserved but unused GPU on most machines which will e-mail you, but don't rely on in only.
-  * Note that the dll machines have typically 10 cards, but "just" 250 GB RAM (DLL6 has only 128 GB). So the expected (maximal) ''mem_free'' requirement for jobs with 1 GPU is 25GB. If your 1-GPU job takes e.g. 80 GB and you submit three such jobs on the same machine, you have effectively blocked the whole machine and seven GPUs remain unused. If you really need to submit more high-memory jobs, send each on different machine.
+  * Note that the dll machines have typically 10 cards, but "just" 250 GB RAM (DLL6 has only 128 GB). So the expected (maximal) ''mem_free'' requirement for jobs with 1 GPU is 25GB. If your 1-GPU job takes e.g. 80 GB and you submit three such jobs on the same machine, you have effectively blocked the whole machine and seven GPUs remain unused. If you really need to submit more high-memory jobs, send each on a different machine.
 ===== How to use cluster =====
@@ Line 42: / Line 42: @@
 ==== Set-up CUDA and CUDNN ====
-You should add the following commands into your ~/.bashrc
+Multiple versions of ''cuda'' can be accessed in ''/opt/cuda''.
+You need to set library path from your ''~/.bashrc'':
-  CUDNN_version=6.0
+  CUDNN_version=7.0
-  CUDA_version=8.0
+  CUDA_version=9.0
-  CUDA_DIR_OPT=/opt/cuda-$CUDA_version
+  CUDA_DIR_OPT=/opt/cuda/$CUDA_version
   if [ -d "$CUDA_DIR_OPT" ] ; then
     CUDA_DIR=$CUDA_DIR_OPT
@@ Line 56: / Line 57: @@
   fi
-When not using Theano, just Tensorflow this can be simplified to ''export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/opt/cuda-8.0/cudnn/6.0/lib64:/opt/cuda-8.0/lib64''. Note that on some machines (dll*, twister*), this is the current default even without setting LD_LIBRARY_PATH, but on other machines (kronos, titan, titan-gpu, iridium) you need to set LD_LIBRARY_PATH explicitly.
+  * When not using Theano, just Tensorflow this can be simplified to ''export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/opt/cuda/9.0/lib64:/opt/cuda/9.0/cudnn/7.0/lib64''.
+  * Note that the ''cudnn'' library is compiled for specific version of ''cuda''. If you need specific version of ''cudnn'', you can look in ''/opt/cuda/$CUDA_version/cudnn/'' whether it is available for a given ''$CUDA_version''.
-TensorFlow 1.5 precompiled binaries need CUDA 9.0, for this you need to
-  export LD_LIBRARY_PATH=/opt/cuda-9.0/lib64/:/opt/cuda/cudnn/7.0/lib64/
-You also need to use ''qsub -q gpu.q@dll[256]'' because only those machines have drivers which support CUDA 9.
-**Testing configuration (so far on twister1 & 2, titan and titan-gpu only)**
-Multiple versions of ''cuda'' can be accessed in ''/opt''.
-System default version for cuda library is configured in ''/etc/ld.so.conf.d/cuda.conf'' as:
-  /opt/cuda/lib64
-  /opt/cuda/extras/CUPTI/lib64
-Actual version used depends on the link in ''/opt''. For example:
-  ls -l /opt
-  ...
-  lrwxrwxrwx 1 root root  8 dub  9 12:30 cuda -> cuda-9.0
-  ...
-This means that the system is using ''cuda 9.0'' by default.
-If system default version does not work for you, you can set library path from your ''~/.bashrc''. Your LD_LIBRARY_PATH should contain:
-  /opt/cuda-$CUDA_version/lib64
-In order to link specific version of ''cudnn'' your LD_LIBRARY_PATH should contain:
-  /opt/cuda-$CUDA_version/cudnn/$CUDNN_version/lib64
-Please check if required CUDNN_version is available for given CUDA_version.
 ==== TensorFlow Environment ====
-Majority people at UFAL use TensorFlow. To start using it you need to create python virtual environment (virtualenv or use Anaconda for it). Into the environment you must place TensorFlow. The TF is either in CPU or GPU version.
+Many people at UFAL use TensorFlow. To start using it it is recommended to create a [[python|Python virtual environment]] (or use Anaconda for it). Into the environment you must place TensorFlow. The TF is either in CPU or GPU version.
   pip install tensorflow
@@ Line 109: / Line 74: @@
 And then you can activate your environment:
-  source activate tf1
+  source activate tf18
-  source activate tf1cpu
+  source activate tf18cpu
-This environment have TensorFlow 1.0 and all necessary requirements for NeuralMonkey.
+This environment have TensorFlow 1.8.0 and all necessary requirements for NeuralMonkey.
 ==== Pytorch Environment ====
@@ Line 126: / Line 91: @@
 As an alternative to ''qsub'', you can use /home/bojar/tools/shell/qsubmit
-  qsubmit --gpumem=2G --queue="gpu.q" WHAT_SHOULD_BE_RUN
+  qsubmit --gpumem=2G --queue="gpu-ms.q" WHAT_SHOULD_BE_RUN
 It is recommended to use priority lower than the default -100 if you are not rushing for the results and don't need to leap over your colleagues jobs.
@@ Line 202: / Line 167: @@
 ==== Individual acquisitions: NVIDIA Academic Hardware Grants ====
-There is an easy way to get one high-end GPU: [[https://developer.nvidia.com/academic_gpu_seeding|ask NVIDIA for an Academic Hardware Grant]]. All it takes is writing a short grant application (at most ~2 hrs of work from scratch; if you have a GAUK, ~15 minutes of copy-pasting). Due to the GPU housing issues (mainly rack space and cooling), Milan F. said we should request the Tesla-line cards (2017 check with Milan about this issue). If you want to have a look at an application, feel free to ask at hajicj@ufal.mff.cuni.cz :)
+There is an easy way to get one high-end GPU: [[https://developer.nvidia.com/academic_gpu_seeding|ask NVIDIA for an Academic Hardware Grant]]. All it takes is writing a short grant application (at most ~2 hrs of work from scratch; if you have a GAUK, ~15 minutes of copy-pasting). Due to the GPU housing issues (mainly rack space and cooling), Milan F. said we should request the Tesla-line cards (2017; check with Milan about this issue). If you want to have a look at an application, feel free to ask at hajicj@ufal.mff.cuni.cz :)
 Take care, however, to coordinate the grant applications a little, so that not too many arrive from UFAL within a short time: these grants are explicitly //not// intended to build GPU clusters, they are "seeding" grants meant for researchers to try out GPUs (and fall in love with them, and buy a cluster later). If you are planning to submit the hardware grant, have submitted one, or have already been awarded one, please add yourself here.

[ Back to the navigation ] [ Back to the content ]

Institute of Formal and Applied Linguistics Wiki

Differences