Comparison of deep learning software

The following tables compare notable software frameworks, libraries, and computer programs for deep learning applications.

Deep learning software by name

Initial Software Open OpenMP CUDA

Software Creator Platform Written in Interface OpenCL support
release license[a] source support support

Jason Dai Apache

BigDL 2016 Yes Apache Spark Scala Scala, Python No
(Intel) 2.0
Berkeley
Vision and Linux, macOS, Python,
Caffe 2013 BSD Yes C++ Yes Under development[4] Yes
Learning Windows[3] MATLAB, C++
Center
Preferred
Chainer 2015 BSD Yes Linux, macOS Python Python No No Yes
Networks
Skymind
engineering
Linux, macOS,
team; Java, Scala,
Apache Windows,
Deeplearning4j Deeplearning4j 2014
2.0
Yes
Android (Cross-
C++, Java Clojure, Python Yes No[8] Yes[9][10]
community; (Keras), Kotlin
platform)
originally
Adam Gibson
Boost
Dlib Davis King 2002 Software Yes Cross-platform C++ C++, Python Yes No Yes
License
Linux, MacOS,
MIT Windows
Flux Mike Innes 2017 Yes Julia Julia Yes
license (Cross-
platform)
Intel Data
Apache Linux, macOS, C++,
Analytics C++, Python,
Intel 2015 License Yes Windows on Python, Yes No No
Acceleration Java[14]
Library
2.0 Intel CPU[14] Java

Intel Math
Linux, macOS, C/C++,
Kernel Library
Intel 2017 Proprietary No Windows on DPC++, C[17] Yes[18] No No
2017 [15] and
Intel CPU[16] Fortran
later
Apache
Linux, macOS, Only on
Google JAX Google 2018 License Yes Python Python
Windows Linux
2.0
Only if
using Can use Theano,
François MIT Linux, macOS,
Keras 2015 Yes Python Python, R Theano Tensorflow or PlaidML Yes
Chollet license Windows
as as backends
backend
Train with
Parallel
MATLAB + Computing
Deep Learning Toolbox
Toolbox C, C++, and
Linux, macOS,
(formally MathWorks 1992 Proprietary No Java, MATLAB No No generate
Windows
Neural MATLAB CUDA
Network code with
Toolbox) GPU
Coder[23]

Python (Keras),
Windows, C++, Command
Microsoft MIT Linux[29] line,[30]
Microsoft
Cognitive 2016 Yes (macOS via C++ Yes[33] No Yes
Toolkit (CNTK)
Research license[28] BrainScript[31]
Docker on
(.NET on
roadmap)
roadmap[32])
MIT Windows,
ML.NET Microsoft 2018 Yes C#, C++ C#, F#
license Linux, macOS
Linux, macOS,
C++, Python,
Windows,[39][40]
Apache Small C++ Julia, MATLAB,
Apache AWS,
Apache MXNet Software 2015 Yes core JavaScript, Go, Yes No Yes
Foundation
2.0 Android,[41] library R, Scala, Perl,
iOS, Clojure
JavaScript[42]
Neural Linux, macOS, Graphical user
Artelnics 2014 Proprietary No C++ Yes No Yes
Designer Windows interface
GNU
OpenNN Artelnics 2003 Yes Cross-platform C++ C++ Yes No Yes
LGPL
Python,
Apache Linux, macOS, Some OpenCL ICDs
PlaidML Vertex.AI, Intel 2017 Yes C++, Python, C++ ? No
2.0 Windows are not recognized
OpenCL
PyTorch Adam Paszke, 2016 BSD Yes Linux, macOS, Python, C, Python, C++, Yes Via separately Yes
Sam Gross, Windows, C++, Julia, R[47] maintained
Soumith Android[46] CUDA package[48][49][50]
 Chintala,
Gregory
Chanan
(Facebook)
Apache
Apache Apache Linux, macOS, Python, C++,
Software 2015 Yes C++ No Supported in V1.0 Yes
SINGA 2.0 Windows Java
Foundation
Python (Keras),
Linux, macOS, C++, C/C++, Java, On roadmap[55] but
Apache
TensorFlow Google Brain 2015 Yes Windows,[52][53] Python, Go, JavaScript, No already with SYCL[56] Yes
2.0
Android CUDA R,[54] Julia, support
Swift

Université de Under
Theano 2007 BSD Yes Cross-platform Python Python (Keras) Yes Yes
Montréal development[59]

Ronan
Collobert, Linux, macOS, Lua, LuaJIT,[66]
Koray C, utility library Third party
Torch 2002 BSD Yes Windows,[64] C, Lua Yes Yes[70][71]
Kavukcuoglu, for implementations[68][69]
Android,[65] iOS
Clement C++/OpenCL[67]
Farabet
Windows, C++,
Wolfram
Wolfram macOS, Linux, Wolfram Wolfram
Mathematica 2014 Proprietary No Yes No Yes
Research Cloud Language, Language
10[74] and later computing CUDA

Initial Software Open OpenMP CUDA

Software Creator Platform Written in Interface OpenCL support
release license[a] source support support

a. Licenses here are a summary, and are not taken to be complete statements of the licenses. Some libraries may use other libraries internally
under different licenses

Comparison of machine learning model compatibility

Design Compatible with Self-contained Pre-processing and Run-time configuration DNN model Common
Format name
goal other formats DNN Model Post-processing for tuning & calibration interconnect platform
TensorFlow,
Algorithm No / Separate files
Keras, Caffe, No No No No Yes
training in most formats
Torch

Algorithm No / Separate files

ONNX Yes No No No Yes
training in most formats

See also
Comparison of numerical-analysis software
Comparison of statistical packages
Comparison of cognitive architectures
List of datasets for machine-learning research
List of numerical-analysis software

References
1. "Deep Learning — ROCm 4.5.0 documentation" (https://web.archive.org/web/20221205102733/https://rocmdocs.amd.com/en/latest/Deep_le
arning/Deep-learning.html). Archived from the original (https://rocmdocs.amd.com/en/latest/Deep_learning/Deep-learning.html) on 2022-12-
05. Retrieved 2022-09-27.
2. Atilim Gunes Baydin; Barak A. Pearlmutter; Alexey Andreyevich Radul; Jeffrey Mark Siskind (20 February 2015). "Automatic differentiation in
machine learning: a survey". arXiv:1502.05767 (https://arxiv.org/abs/1502.05767) [cs.LG (https://arxiv.org/archive/cs.LG)].
3. "Microsoft/caffe" (https://github.com/Microsoft/caffe). GitHub. 30 October 2021.
4. "Caffe: a fast open framework for deep learning" (https://github.com/BVLC/caffe). July 19, 2019 – via GitHub.
5. "Caffe | Model Zoo" (http://caffe.berkeleyvision.org/model_zoo.html). caffe.berkeleyvision.org.
6. GitHub - BVLC/caffe: Caffe: a fast open framework for deep learning. (https://github.com/BVLC/caffe), Berkeley Vision and Learning Center,
2019-09-25, retrieved 2019-09-25
7. Preferred Networks Migrates its Deep Learning Research Platform to PyTorch (https://preferred.jp/en/news/pr20191205/), 2019-12-05,
retrieved 2019-12-27
8. "Support for Open CL · Issue #27 · deeplearning4j/nd4j" (https://github.com/deeplearning4j/nd4j/issues/27). GitHub.
9. "N-Dimensional Scientific Computing for Java" (https://web.archive.org/web/20161016094035/http://nd4j.org/gpu_native_backends.html).
Archived from the original (http://nd4j.org/gpu_native_backends.html) on 2016-10-16. Retrieved 2016-02-05.
10. "Comparing Top Deep Learning Frameworks" (https://web.archive.org/web/20171107011631/https://deeplearning4j.org/compare-dl4j-tensorfl
ow-pytorch). Deeplearning4j. Archived from the original (https://deeplearning4j.org/compare-dl4j-tensorflow-pytorch) on 2017-11-07.
Retrieved 2017-10-31.
11. Chris Nicholson; Adam Gibson. "Deeplearning4j Models" (https://web.archive.org/web/20170211020819/https://deeplearning4j.org/model-zo
o). Archived from the original (http://deeplearning4j.org/model-zoo) on 2017-02-11. Retrieved 2016-03-02.
12. Deeplearning4j. "Deeplearning4j on Spark" (https://web.archive.org/web/20170713012632/https://deeplearning4j.org/spark). Deeplearning4j.
Archived from the original (http://deeplearning4j.org/spark) on 2017-07-13. Retrieved 2016-09-01.
13. "Metalhead" (https://github.com/FluxML/Metalhead.jl). FluxML. 29 October 2021.
14. "Intel® Data Analytics Acceleration Library (Intel® DAAL)" (https://software.intel.com/en-us/intel-daal). software.intel.com. November 20,
2018.
15. "Intel® Math Kernel Library Release Notes and New Features" (https://www.intel.com/content/www/us/en/developer/articles/release-notes/int
el-math-kernel-library-release-notes-and-new-features.html). Intel.
16. "Intel® Math Kernel Library (Intel® MKL)" (https://software.intel.com/en-us/mkl). software.intel.com. September 11, 2018.
17. "Deep Neural Network Functions" (https://software.intel.com/en-us/mkl-developer-reference-c-deep-neural-network-functions).
software.intel.com. May 24, 2019.
18. "Using Intel® MKL with Threaded Applications" (https://software.intel.com/en-us/articles/intel-math-kernel-library-intel-mkl-using-intel-mkl-with-
threaded-applications). software.intel.com. June 1, 2017.
19. "Intel® Xeon Phi™ Delivers Competitive Performance For Deep Learning—And Getting Better Fast" (https://software.intel.com/en-us/articles/i
ntel-xeon-phi-delivers-competitive-performance-for-deep-learning-and-getting-better-fast). software.intel.com. March 21, 2019.
20. "Applications - Keras Documentation" (https://keras.io/applications/). keras.io.
21. "Is there RBM in Keras? · Issue #461 · keras-team/keras" (https://github.com/keras-team/keras/issues/461). GitHub.
22. "Does Keras support using multiple GPUs? · Issue #2436 · keras-team/keras" (https://github.com/keras-team/keras/issues/2436). GitHub.
23. "GPU Coder - MATLAB & Simulink" (https://www.mathworks.com/products/gpu-coder.html). MathWorks. Retrieved 13 November 2017.
24. "Automatic Differentiation Background - MATLAB & Simulink" (https://www.mathworks.com/help/deeplearning/ug/deep-learning-with-automati
c-differentiation-in-matlab.html). MathWorks. September 3, 2019. Retrieved November 19, 2019.
25. "Neural Network Toolbox - MATLAB" (https://www.mathworks.com/products/neural-network.html). MathWorks. Retrieved 13 November 2017.
26. "Deep Learning Models - MATLAB & Simulink" (https://www.mathworks.com/solutions/deep-learning/models.html). MathWorks. Retrieved
13 November 2017.
27. "Parallel Computing Toolbox - MATLAB" (https://www.mathworks.com/products/parallel-computing.html). MathWorks. Retrieved 13 November
2017.
28. "CNTK/LICENSE.md at master · Microsoft/CNTK" (https://github.com/Microsoft/CNTK/blob/master/LICENSE.md). GitHub.
29. "Setup CNTK on your machine" (https://github.com/Microsoft/CNTK/wiki/Setup-CNTK-on-your-machine). GitHub.
30. "CNTK usage overview" (https://github.com/Microsoft/CNTK/wiki/CNTK-usage-overview). GitHub.
31. "BrainScript Network Builder" (https://github.com/Microsoft/CNTK/wiki/BrainScript-Network-Builder). GitHub.
32. ".NET Support · Issue #960 · Microsoft/CNTK" (https://github.com/Microsoft/CNTK/issues/960). GitHub.
33. "How to train a model using multiple machines? · Issue #59 · Microsoft/CNTK" (https://github.com/Microsoft/CNTK/issues/59#issuecomment-
178104505). GitHub.
34. "Prebuilt models for image classification · Issue #140 · microsoft/CNTK" (https://github.com/microsoft/CNTK/issues/140). GitHub.
35. "CNTK - Computational Network Toolkit" (http://www.cntk.ai/). Microsoft Corporation.
36. "Restricted Boltzmann Machine with CNTK #534" (https://github.com/Microsoft/CNTK/issues/534). GitHub, Inc. 27 May 2016. Retrieved
30 October 2023.
37. "Multiple GPUs and machines" (https://github.com/Microsoft/CNTK/wiki/Multiple-GPUs-and-machines). Microsoft Corporation.
38. "Disclaimer" (https://github.com/Microsoft/CNTK#disclaimer). CNTK TEAM. 6 November 2021.
39. "Releases · dmlc/mxnet" (https://github.com/dmlc/mxnet/releases). Github.
40. "Installation Guide — mxnet documentation" (https://mxnet.readthedocs.io/en/latest/how_to/build.html#building-on-windows). Readthdocs.
41. "MXNet Smart Device" (https://web.archive.org/web/20160921205959/http://mxnet.readthedocs.io/en/latest/how_to/smart_device.html).
ReadTheDocs. Archived from the original (https://mxnet.readthedocs.io/en/latest/how_to/smart_device.html) on 2016-09-21. Retrieved
2016-05-19.
42. "MXNet.js" (https://github.com/dmlc/mxnet.js). Github. 28 October 2021.
43. "— Redirecting to mxnet.io" (https://mxnet.readthedocs.io/en/latest/). mxnet.readthedocs.io.
44. "Model Gallery" (https://github.com/dmlc/mxnet-model-gallery). GitHub. 29 October 2022.
45. "Run MXNet on Multiple CPU/GPUs with Data Parallel" (https://mxnet.readthedocs.io/en/latest/how_to/multi_devices.html). GitHub.
46. "PyTorch" (https://pytorch.org/mobile/android/). Dec 17, 2021.
47. "Falbel D, Luraschi J (2023). torch: Tensors and Neural Networks with 'GPU' Acceleration" (https://torch.mlverse.org/). torch.mlverse.org.
Retrieved 2023-11-28.
48. "OpenCL build of pytorch: (in-progress, not useable) - hughperkins/pytorch-coriander" (https://github.com/hughperkins/pytorch-coriander).
July 14, 2019 – via GitHub.
49. "DLPrimitives/OpenCL out of tree backend for pytorch - artyom-beilis/pytorch_dlprim" (https://github.com/artyom-beilis/pytorch_dlprim). Jan
21, 2022 – via GitHub.
50. "OpenCL Support · Issue #488 · pytorch/pytorch" (https://github.com/pytorch/pytorch/issues/488). GitHub.
51. "Restricted Boltzmann Machines (RBMs) in PyTorch" (https://github.com/GabrielBianconi/pytorch-rbm/blob/master/rbm.py). GitHub. 14
November 2022.
52. "Install TensorFlow with pip" (https://www.tensorflow.org/install/pip).
53. "TensorFlow 0.12 adds support for Windows" (https://developers.googleblog.com/2016/11/tensorflow-0-12-adds-support-for-windows.html).
54. Allaire, J.J.; Kalinowski, T.; Falbel, D.; Eddelbuettel, D.; Yuan, T.; Golding, N. (28 September 2023). "tensorflow: R Interface to 'TensorFlow' "
(https://cran.r-project.org/web/packages/tensorflow/). The Comprehensive R Archive Network. Retrieved 30 October 2023.
55. "tensorflow/roadmap.md at master" (https://github.com/tensorflow/tensorflow/blob/master/tensorflow/docs_src/about/roadmap.md). GitHub.
January 23, 2017. Retrieved May 21, 2017.
56. "OpenCL support" (https://github.com/tensorflow/tensorflow/issues/22). GitHub.
57. "TensorFlow" (https://www.tensorflow.org/). TensorFlow.
58. "Models and examples built with TensorFlow" (https://github.com/tensorflow/models). July 19, 2019 – via GitHub.
59. "Using the GPU: Theano 0.8.2 documentation" (https://web.archive.org/web/20170401163303/http://deeplearning.net/software/theano/tutoria
l/using_gpu.html). Archived from the original (http://deeplearning.net/software/theano/tutorial/using_gpu.html) on 2017-04-01. Retrieved
2016-01-21.
60. "gradient – Symbolic Differentiation — Theano 1.0.0 documentation" (http://deeplearning.net/software/theano/library/gradient.html).
deeplearning.net.
61. "Automatic vs. Symbolic differentiation" (https://groups.google.com/d/msg/theano-users/mln5g2IuBSU/gespG36Lf_QJ).
62. "Recipes/modelzoo at master · Lasagne/Recipes" (https://github.com/Lasagne/Recipes/tree/master/modelzoo). GitHub.
63. "Using multiple GPUs — Theano 1.0.0 documentation" (http://deeplearning.net/software/theano/tutorial/using_multi_gpu.html).
deeplearning.net.
64. "torch/torch7" (https://github.com/torch/torch7). July 18, 2019 – via GitHub.
65. "GitHub - soumith/torch-android: Torch-7 for Android" (https://github.com/soumith/torch-android). GitHub. 13 October 2021.
66. "Torch7: A MATLAB-like Environment for Machine Learning" (http://ronan.collobert.com/pub/matos/2011_torch7_nipsw.pdf) (PDF).
67. "GitHub - jonathantompson/jtorch: An OpenCL Torch Utility Library" (https://github.com/jonathantompson/jtorch). GitHub. 18 November 2020.
68. "Cheatsheet" (https://github.com/torch/torch7/wiki/Cheatsheet#opencl). GitHub.
69. "cltorch" (https://github.com/hughperkins/distro-cl). GitHub.
70. "Torch CUDA backend" (https://github.com/torch/cutorch). GitHub.
71. "Torch CUDA backend for nn" (https://github.com/torch/cunn). GitHub.
72. "Autograd automatically differentiates native Torch code: twitter/torch-autograd" (https://github.com/twitter/torch-autograd). July 9, 2019 – via
GitHub.
73. "ModelZoo" (https://github.com/torch/torch7/wiki/ModelZoo). GitHub.
74. "Launching Mathematica 10" (https://blog.wolfram.com/2014/07/09/launching-mathematica-10-with-700-new-functions-and-a-crazy-amount-of
-rd). Wolfram.
75. "Wolfram Neural Net Repository of Neural Network Models" (http://resources.wolframcloud.com/NeuralNetRepository).
resources.wolframcloud.com.
76. "Parallel Computing—Wolfram Language Documentation" (https://reference.wolfram.com/language/guide/ParallelComputing.html.en).
reference.wolfram.com.
77. "Deep Learning — ROCm 4.5.0 documentation" (https://web.archive.org/web/20221205102733/https://rocmdocs.amd.com/en/latest/Deep_le
arning/Deep-learning.html). Archived from the original (https://rocmdocs.amd.com/en/latest/Deep_learning/Deep-learning.html) on 2022-12-
05. Retrieved 2022-09-27.

Retrieved from "https://en.wikipedia.org/w/index.php?title=Comparison_of_deep_learning_software&oldid=1254068915"