Differences

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

--- gpu [2017/03/16 16:30]
kocmanek [Servers with GPU units]
+++ gpu [2018/03/21 15:29]
ufal [Servers with GPU units]
@@ Line 4: / Line 4: @@
 ===== Servers with GPU units =====
+GPU cluster ''gpu.q'' at Malá Strana:
-| machine | GPU; [[https://en.wikipedia.org/wiki/CUDA#Supported_GPUs|Capability]] [cc]  | cores | GPU RAM | Comment |
+| machine | GPU type | GPU driver version | [[https://en.wikipedia.org/wiki/CUDA#GPUs_supported|cc]] | GPU cnt | GPU RAM (MB) | machine RAM (GB)|
-| titan-gpu  | GeForce GTX Titan Z; cc3.5 | 2 | 6 GB each core |        |
+| dll1 | GeForce GTX 1080 | 384.69 | 6.1 | 8 | 8114 | 250 |
-| twister1   | Tesla K40c; cc3.5          | 1 | 12 GB          |        |
+| dll2 | GeForce GTX 1080 | 387.34 | 6.1 | 7 | 8114 | 250 |
-| twister2   | Tesla K40c; cc3.5          | 1 | 12 GB          |        |
+| dll3 | GeForce GTX 1080 Ti | 375.66 | 6.1 | 9 | 11172 | 250 |
-| kronos-dev | Tesla K40c; cc3.5          | 1 | 12 GB          |        |
+| dll4 | GeForce GTX 1080 Ti | 375.66 | 6.1 | 10 | 11172 | 250 |
-| iridium    | Quadro K2000; cc3.0        | 1 | 2 GB           |        |
+| dll5 | GeForce GTX 1080 Ti | 384.69 | 6.1 | 10 | 11172 | 250 |
-| victoria   | GeForce GT 630; cc3.0      | 1 | 2 GB           | Ondrej Bojar's desktop machine |
+| dll6 | GeForce GTX 1080 Ti | 384.69 | 6.1 | 9 | 11172 | 122 |
-| arc        | GeForce GT 630; cc3.0      | 1 | 2 GB           | Ales's desktop machine |
+| titan | GeForce GTX 1080 | 381.22 | 6.1 | 1 | 8114 | 31 |
-| athena     | GeForce GTX 1080; cc6.1    | 1 | 8 GB           | Tom's desktop machine |
+| twister1 | Tesla K40c | 367.48 | 3.5 | 1 | 11439 | 47 |
-| dll1     | GeForce GTX 1080; cc6.1    | 8 | 8 GB each core |  |
+| twister2 | Tesla K40c | 367.48 | 3.5 | 1 | 11439 | 47 |
-| dll2     | GeForce GTX 1080; cc6.1    | 8 | 8 GB each core |  |
+| titan-gpu | GeForce GTX TITAN Z | 381.22 | 3.5 | 2 | 6082 | 31 |
+| kronos | GeForce GTX 1080 Ti | 384.81 | 6.1 | 1 | 11172 | 125 |
+| iridium | Quadro K2000 | 367.48 | 3.0 | 1 | 1998 | 504 |
-not used at the moment: GeForce GTX 570 (from twister2)
-All machines have CUDA8.0 and should support both Theano and TensorFlow.
-Summary of future plans:
+Desktop machines:
-  * Current Troja servers won't get any GPUs (the only option would be [[http://www.czc.cz/hp-quadro-k1200-4gb/171662/produkt?ppcbee-adtext-variant=Produkt%3B+kategorie+%2B+cena%3B+Pobo%C4%8Dky&gclid=CKbKkbrWrswCFQUq0wodHDELCw|Quadro K1200 4GB]], horribly cost-inefficient)
+| machine                    | GPU type | [[https://en.wikipedia.org/wiki/CUDA#GPUs_supported|cc]] | GPUs | GPU RAM | Comment |
-  * The old Quadro K2000 we have is a much more low end piece, so we can't test is in Troja.
+| victoria; arc              | GeForce GT 630   | cc3.0 |  1 |  2 GB | desktop machine |
-  * There is MetaCentrum which also has GPUs, so testing can be done there.
+| athena                     | GeForce GTX 1080 | cc6.1 |  1 |  8 GB | Tom's desktop machine |
-  * It is impossible (wasteful in terms of space and forbidden by a dean regulation) to put non-rack machines to our servers rooms. So we won't be buying GeForce GTX 1080 (~20000CZK, out of stock now), for a non-rack machine since we most likely don't have any available.
-  * Yes, there are grant applications under review which include rack machines with GPUs, e.g. 5x2 or something like that; more will be known in 2017.
+Not used at the moment: GeForce GTX 570 (from twister2)
+All machines have CUDA8.0 and should support both Theano and TensorFlow.
-=== Individual acquisitions: NVIDIA Academic Hardware Grants ==
+[[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]]
-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. 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.
+===== Rules =====
+  * 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>
+  * 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.
+  * 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 ''$CUDA_VISIBLE_DEVICES'' will not be set correctly. E.g. <code>qrsh -q gpu.q -l gpu=1,gpu_ram=2G -pty yes bash</code>In general: don't reserve a GPU (as described above) without actually using it for longer time. (E.g. try separating steps which need GPU and steps which do not and execute those separately on our GPU resp. CPU cluster.) Ondřej Bojar has a script /home/bojar/tools/servers/watch_gpus for watching reserved but unused GPU on most machines which will e-mail you, but don't rely on in only.
+  * Note that the dll machines have typically 10 cards, but "just" 250 GB RAM (DLL6 has only 128 GB). So the expected (maximal) ''mem_free'' requirement for jobs with 1 GPU is 25GB. If your 1-GPU job takes e.g. 80 GB and you submit three such jobs on the same machine, you have effectively blocked the whole machine and seven GPUs remain unused. If you really need to submit more high-memory jobs, send each on different machine.
-Known NVIDIA Academic Hardware Grants:
+===== How to use cluster =====
-  * Ondřej Plátek - granted (2015)
+==== Set-up CUDA and CUDNN ====
-  * Jan Hajič jr. - granted (early 2016)
-  * Jindra Helcl - planning to apply (fall 2016)
+You should add the following commands into your ~/.bashrc
+  CUDNN_version=6.0
+  CUDA_version=8.0
+  CUDA_DIR_OPT=/opt/cuda-$CUDA_version
+  if [ -d "$CUDA_DIR_OPT" ] ; then
+    CUDA_DIR=$CUDA_DIR_OPT
+    export CUDA_HOME=$CUDA_DIR
+    export THEANO_FLAGS="cuda.root=$CUDA_HOME,device=gpu,floatX=float32"
+    export PATH=$PATH:$CUDA_DIR/bin
+    export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:$CUDA_DIR/cudnn/$CUDNN_version/lib64:$CUDA_DIR/lib64
+    export CPATH=$CUDA_DIR/cudnn/$CUDNN_version/include:$CPATH
+  fi
-===== Performance tests =====
+When not using Theano, just Tensorflow this can be simplified to ''export LD_LIBRARY_PATH=$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.
-* [[http://www.trustedreviews.com/nvidia-geforce-gtx-1080-review-performance-benchmarks-and-conclusion-page-2| 980 vs 1080 vs Titan X (not the Titan Z we have)]]
+TensorFlow 1.5 precompiled binaries need CUDA 9.0, for this you need to
-In the following table is the experiment conducted by Tom Kocmi. You can replicate experiment: /a/merkur3/kocmanek/Projects/GPUBenchmark (you will need to prepare environment of TensorFlow11 or use my ANACONDA)
+  export LD_LIBRARY_PATH=/opt/cuda-9.0/lib64/:/opt/cuda/cudnn/7.0/lib64/
-I am preparing department-wide benchmark, but meanwhile the results for different experiment:
+You also need to use ''qsub -q gpu.q@dll[256]'' because only those machines have drivers which support CUDA 9.
- * Athena (GTX 1080) - 2 hodiny 32 minut
- * Twister (Tesla K40c) - 6 hodin 46 minut
-| machine | Setup; CPU/GPU; [[https://en.wikipedia.org/wiki/CUDA#Supported_GPUs|Capability]] [cc] | Walltime | Note |
+==== TensorFlow Environment ====
-| athena     | GeForce GTX 1080; cc6.1            |    9:55:58 | Tom's desktop  |
-| dll2       | (2 GPU) GeForce GTX 1080; cc6.1    |   10:19:40 | with CUDA_VISIBLE_DEVICES=0 |
-| dll1       | (2 GPU) GeForce GTX 1080; cc6.1    |   12:34:34 | Probably only one GPU was used |
-| dll2       | (2 GPU) GeForce GTX 1080; cc6.1    |   13:01:05 | Only one GPU was used |
-| titan-gpu  | (2 GPU) GeForce GTX Titan Z; cc3.5 |   16:05:24 | Probably only one GPU was used |
-| kronos-dev | Tesla K40c; cc3.5                  |   22:41:01 |  |
-| twister2   | Tesla K40c; cc3.5                  |   22:43:10 |  |
-| twister1   | Tesla K40c; cc3.5                  |   24:19:45 |  |
-| helena1    | 16x cores CPU                      |   46:33:14 |  |
-| belzebub   | 16x cores CPU                      |   52:36:56 |  |
-| iridium    | Quadro K2000; cc3.0                |   59:47:58 |  |
-| helena7    | 8x cores CPU                       |   60:39:17 |  |
-| arc        | GeForce GT 630; cc3.0              |  103:42:30 | (approximated after 66 hours) |
-| lucifer4   | 8x cores CPU                       |  134:41:22 |  |
-| victoria   | GeForce GT 630; cc3.0              |        --- | never run, same GPU as Arc |
+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.
-A comparison with Ondrej's small data set:
+  pip install tensorflow
-  * dll2 (2xGPU) takes 13m for one reporting period
+  pip install tensorflow-gpu
-  * achilles2 (4xCPU with 8 CPUs reserved) takes 24m for one reporting period
+You can use prepared environment by adding into your ~/.bashrc
+  export PATH=/a/merkur3/kocmanek/ANACONDA/bin:$PATH
-===== Installed toolkits =====
+And then you can activate your environment:
-//This should mention where each interesting toolkit lives (on a particular machine).//
+  source activate tf1
+  source activate tf1cpu
-==== TensorFlow ====
+This environment have TensorFlow 1.0 and all necessary requirements for NeuralMonkey.
-[[https://redmine.ms.mff.cuni.cz/projects/mmmt/repository/revisions/6a064187fc6959db9b77cf2d5350c5f4918a8067/entry/prepare_env.sh|This script]] installs TensorFlow 0.7.1 (and all other dependencies we need for Multimodal Translation) into `tf' and `tf-gpu' virtual environments. The GPU environment can be loaded by calling <code>source tf-gpu/bin/activate-gpu</code>
+==== Pytorch Environment ====
-OP: I created [[https://gist.github.com/oplatek/323b63b8f116cd3d78c0f492f78cc289|script]] which install Tensorflow 0.8 and test it if it uses GPU. TF is installed into `user` or `global` installation either for `python3.4` or `python2.7`
+If you want to use pytorch, there is a ready-made environment in
-=== Select GPU device ===
+  /home/hajicj/anaconda3/envs/pytorch/bin
+It does rely on the CUDA and CuDNN setup above.
-Use variable CUDA_VISIBLE_DEVICES to constrain tensorflow to compute only on the selected one. For the use of first GPU use:
+==== Using cluster ====
-<code>export CUDA_VISIBLE_DEVICES=0</code>
-To list available devices, use:
+As an alternative to ''qsub'', you can use /home/bojar/tools/shell/qsubmit
-<code>/opt/cuda/samples/1_Utilities/deviceQuery/deviceQuery | grep ^Device</code>
-===== Basic commands =====
+  qsubmit --gpumem=2G --queue="gpu.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.
+==== Basic commands ====
   lspci
@@ Line 111: / Line 116: @@
   /usr/local/cuda/samples/1_Utilities/deviceQuery/deviceQuery
     # shows CUDA capability etc.
+  ssh dll1; ~popel/bin/gpu_allocations
+    # who occupies which card on a given machine
+=== Select GPU device ===
+The variable CUDA_VISIBLE_DEVICES constrains tensorflow and other toolkits to compute only on the selected GPUs. **Do not set this variable yourself** (unless debugging SGE), it is set for you automatically by SGE if you ask for some GPUs (see above).
+To list available devices, use:
+  /opt/cuda/samples/1_Utilities/deviceQuery/deviceQuery | grep ^Device
+===== Performance tests =====
+* [[http://www.trustedreviews.com/nvidia-geforce-gtx-1080-review-performance-benchmarks-and-conclusion-page-2| 980 vs 1080 vs Titan X (not the Titan Z we have)]]
+In the following table is the experiment conducted by Tom Kocmi. You can replicate experiment: /a/merkur3/kocmanek/Projects/GPUBenchmark (you will need to prepare environment of TensorFlow11 or use my ANACONDA). The benchmark uses 2GB model of seq2seq machine translation in Neural Monkey (De > EN). If not specified, the benchmark had an access only to one GPU.
+| machine | Setup; CPU/GPU; [[https://en.wikipedia.org/wiki/CUDA#Supported_GPUs|Capability]] [cc] | Walltime | Note |
+| athena     | GeForce GTX 1080; cc6.1            |    9:55:58 | Tom's desktop  |
+| dll2       | GeForce GTX 1080; cc6.1            |   10:19:40 |  |
+| titan      | GeForce GTX 1080 Ti                |   10:45:11 | (new result with correct CUDA version) |
+| dll1       | (2 GPU) GeForce GTX 1080; cc6.1    |   12:34:34 | Probably only one GPU was used |
+| dll2       | GeForce GTX 1080; cc6.1            |   13:01:05 | Only one GPU was used |
+| titan-gpu  | (2 GPU) GeForce GTX Titan Z; cc3.5 |   16:05:24 | Probably only one GPU was used |
+| kronos-dev | Tesla K40c; cc3.5                  |   22:41:01 |  |
+| twister2   | Tesla K40c; cc3.5                  |   22:43:10 |  |
+| twister1   | Tesla K40c; cc3.5                  |   24:19:45 |  |
+| helena1    | 16x cores CPU                      |   46:33:14 |  |
+| belzebub   | 16x cores CPU                      |   52:36:56 |  |
+| iridium    | Quadro K2000; cc3.0                |   59:47:58 |  |
+| helena7    | 8x cores CPU                       |   60:39:17 |  |
+| arc        | GeForce GT 630; cc3.0              |  103:42:30 | (approximated after 66 hours) |
+| lucifer4   | 8x cores CPU                       |  134:41:22 |  |
+=== Second Benchmark ===
+The previous benchmark only compares the speed of processing units within the GPUs and do not take into account the size of memory. Therefore I have conducted another benchmark, this time for each graphic card I have increased the batch size as much as possible so the model still could fit into the GPU (the previous benchmark model had batch size 20). This way the results should be more representative of the power for each GPU.
+| GPU; Cuda capability       | GPU RAM |  Walltime | Batch size | Machine |
+| Tesla K40c; cc3.5          |   12 GB |           |        |  |
+| GeForce GTX 1080 Ti; cc6.1 |   11 GB |  00:55:56 |       2300 | dll5 |
+| GeForce GTX 1080; cc6.1    |    8 GB |  01:10:57 |       1700 | dll1 |
+| GeForce GTX Titan Z; cc3.5 |    6 GB |  02:20:47 |       1100 | titan-gpu |
+| Quadro K2000; cc3.0        |    2 GB |  28:15:26 |         50 | iridium |
 ===== Links =====
@@ Line 119: / Line 168: @@
 GPU specs for those GPUs we have:
   * [[http://www.nvidia.com/content/PDF/kepler/Tesla-K40-Active-Board-Spec-BD-06949-001_v03.pdf|Tesla K40c]]
+==== 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 :)
+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.
+Known NVIDIA Academic Hardware Grants:
+  * Ondřej Plátek - granted (2015)
+  * Jan Hajič jr. - granted (early 2016)

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

Institute of Formal and Applied Linguistics Wiki

Differences