Differences
This shows you the differences between two versions of the page.
Both sides previous revision Previous revision Next revision | Previous revision Next revision Both sides next revision | ||
gpu [2017/05/16 10:48] kocmanek [Performance tests] |
gpu [2021/12/15 17:53] kasner Update CUDA version |
||
---|---|---|---|
Line 4: | Line 4: | ||
===== Servers with GPU units ===== | ===== Servers with GPU units ===== | ||
+ | GPU cluster '' | ||
- | | machine | + | | machine | GPU type | GPU driver version | [[https:// |
- | | titan | + | | dll1 | |
- | | titan-gpu | + | | dll3 | |
- | | twister1 | + | | dll4 | |
- | | twister2 | + | | dll5 | |
- | | kronos-dev | + | | dll6 | NVIDIA RTX A4000 |
- | | iridium | + | | dll7 | GeForce |
- | | victoria | + | | dll8 | |
- | | arc | GeForce GT 630; cc3.0 | + | | dll9 | GeForce |
- | | athena | + | | dll10 | GeForce |
- | | dll1 | GeForce | + | |
- | | dll2 | GeForce | + | |
- | not used at the moment: GeForce GTX 570 (from twister2) | + | GPU cluster '' |
- | All machines have CUDA8.0 and should support both Theano and TensorFlow. | + | |
- | Summary of future plans: | + | | machine | GPU type | GPU driver version | [[https://en.wikipedia.org/wiki/CUDA# |
- | * Current Troja servers won't get any GPUs (the only option would be [[http://www.czc.cz/hp-quadro-k1200-4gb/171662/ | + | | tdll1 | |
- | | + | | tdll2 | Quadro P5000 | 455.23.05 | 6.1 | 8 | 17 | 245.0 | |
- | | + | | tdll3 | Quadro P5000 | 455.23.05 | 6.1 | 8 | 17 | 245.0 | |
- | | + | | tdll4 | Quadro P5000 | 455.23.05 | 6.1 | 8 | 17 | 245.0 | |
- | | + | | tdll5 | Quadro P5000 | |
+ | Desktop machines: | ||
+ | | machine | ||
+ | | victoria; arc | GeForce GT 630 | cc3.0 | 1 | 2 GB | desktop machine | | ||
+ | | athena | ||
- | === Individual acquisitions: | + | Multiple versions of CUDA library are accessible on each machine together with cudnn. Theano and TensorFlow is supported. |
- | 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, | + | [[http://ufaladm2.ufal.hide.ms.mff.cuni.cz/munin/ufal.hide.ms.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 " | ||
- | Known NVIDIA Academic Hardware Grants: | + | ===== Rules ===== |
+ | * First, read [[internal:Linux network]] and [[: | ||
+ | * 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 '' | ||
+ | * **Note that you need to use '' | ||
+ | * Always specify the number of GPU cards (e.g. '' | ||
+ | * If you need more than one GPU card (on a single machine), always require at least as many CPU cores ('' | ||
+ | * For interactive jobs, you can use '' | ||
+ | * Note that the dll machines have typically 10 cards, but " | ||
+ | * If you know an approximate runtime of your job, please specify it with '' | ||
- | * Ondřej Plátek - granted (2015) | ||
- | * Jan Hajič jr. - granted (early 2016) | ||
- | * Jindra Helcl - planning to apply (fall 2016) | ||
+ | ===== How to use cluster ===== | ||
- | | + | ==== Set-up CUDA and CUDNN ==== |
+ | |||
+ | Multiple versions of '' | ||
+ | |||
+ | You need to set library path from your '' | ||
+ | |||
+ | CUDA_version=11.1 | ||
+ | CUDNN_version=8.0 | ||
+ | CUDA_DIR_OPT=/ | ||
+ | if [ -d " | ||
+ | CUDA_DIR=$CUDA_DIR_OPT | ||
+ | export CUDA_HOME=$CUDA_DIR | ||
+ | export THEANO_FLAGS=" | ||
+ | export PATH=$PATH: | ||
+ | export LD_LIBRARY_PATH=$LD_LIBRARY_PATH: | ||
+ | export CPATH=$CUDA_DIR/ | ||
+ | fi | ||
+ | |||
+ | * When not using Theano, just Tensorflow this can be simplified to '' | ||
+ | * Note that the '' | ||
- | ===== How to use cluster ===== | ||
- | In this section will be explained how to use cluster properly. | ||
==== TensorFlow Environment ==== | ==== TensorFlow Environment ==== | ||
- | Majority | + | Many people at UFAL use TensorFlow. To start using it it is recommended |
pip install tensorflow | pip install tensorflow | ||
Line 60: | Line 84: | ||
And then you can activate your environment: | And then you can activate your environment: | ||
- | source activate | + | source activate |
- | source activate | + | source activate |
- | 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. |
- | ==== Using cluster | + | ==== Pytorch Environment |
- | Rule number one, always use the GPU queue (never log in machine by ssh). Always use qsub or qsubmit with proper arguments. | + | If you want to use pytorch, there is a ready-made environment |
- | For testing and using the cluster interactively you can use qrsh (this should not be used for long running experiments since the console is not closed on the end of the experiment). Following command will assign you a GPU and creates interactive console. | + | / |
- | + | ||
- | qrsh -q gpu.q -l gpu=1 -pty yes bash | + | |
| | ||
- | For running experiments you must use qsub command: | + | It does rely on the CUDA and CuDNN setup above. |
- | qsub -q gpu.q -l gpu=1, | + | ==== Using cluster ==== |
- | + | ||
- | Cleaner way to use cluster is with / | + | As an alternative |
- | qsubmit --gpumem=2G --queue=" | + | qsubmit --gpumem=2G --queue=" |
| | ||
- | It is recommended to use priority -100 if you are not rushing for the results and don't need to leap over your colleagues jobs. | + | It is recommended to use priority |
==== Basic commands ==== | ==== Basic commands ==== | ||
Line 91: | Line 113: | ||
watch nvidia-smi | watch nvidia-smi | ||
# For monitoring GPU activity in a separate terminal (thanks to Jindrich Libovicky for this!) | # For monitoring GPU activity in a separate terminal (thanks to Jindrich Libovicky for this!) | ||
+ | # You can also use nvidia-smi -l TIME | ||
nvcc --version | nvcc --version | ||
# this should tell CUDA version | # this should tell CUDA version | ||
Line 99: | Line 122: | ||
/ | / | ||
# shows CUDA capability etc. | # shows CUDA capability etc. | ||
+ | ssh dll1; ~popel/ | ||
+ | # who occupies which card on a given machine | ||
| | ||
=== Select GPU device === | === Select GPU device === | ||
- | Use variable CUDA_VISIBLE_DEVICES | + | The variable CUDA_VISIBLE_DEVICES |
- | export CUDA_VISIBLE_DEVICES=0 | + | |
To list available devices, use: | To list available devices, use: | ||
Line 112: | Line 136: | ||
* [[http:// | * [[http:// | ||
- | In the following table is the experiment conducted by Tom Kocmi. You can replicate experiment: / | + | In the following table is the experiment conducted by Tom Kocmi. You can replicate experiment: / |
| machine | Setup; CPU/GPU; [[https:// | | machine | Setup; CPU/GPU; [[https:// | ||
| athena | | athena | ||
- | | dll2 | + | | dll2 | GeForce GTX 1080; cc6.1 | |
- | | titan | GeForce GTX 1080 Ti | 11:41:08 | | | + | | titan | GeForce GTX 1080 Ti | 10:45:11 | (new result with correct CUDA version) |
| dll1 | (2 GPU) GeForce GTX 1080; cc6.1 | | | dll1 | (2 GPU) GeForce GTX 1080; cc6.1 | | ||
- | | dll2 | + | | twister2 |
+ | | dll2 | GeForce GTX 1080; cc6.1 | | ||
| titan-gpu | | titan-gpu | ||
| kronos-dev | Tesla K40c; cc3.5 | | | kronos-dev | Tesla K40c; cc3.5 | | ||
Line 130: | Line 155: | ||
| arc | GeForce GT 630; cc3.0 | 103:42:30 | (approximated after 66 hours) | | | arc | GeForce GT 630; cc3.0 | 103:42:30 | (approximated after 66 hours) | | ||
| lucifer4 | | lucifer4 | ||
- | | victoria | ||
+ | === 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 | ||
+ | | 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 | | ||
+ | | Quadro P5000 | ||
+ | | 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 ===== | ===== Links ===== | ||
Line 141: | Line 175: | ||
GPU specs for those GPUs we have: | GPU specs for those GPUs we have: | ||
* [[http:// | * [[http:// | ||
+ | |||
+ | ==== Individual acquisitions: | ||
+ | |||
+ | There is an easy way to get one high-end GPU: [[https:// | ||
+ | |||
+ | 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 " | ||
+ | |||
+ | Known NVIDIA Academic Hardware Grants: | ||
+ | |||
+ | * Ondřej Plátek - granted (2015) | ||
+ | * Jan Hajič jr. - granted (early 2016) | ||
+ | |||
+ | |||
+ |