About: DeepSpeed

An Entity of Type: software, from Named Graph: http://dbpedia.org, within Data Space: dbpedia.org

DeepSpeed is an open source deep learning optimization library for PyTorch. The library is designed to reduce computing power and memory use and to train large distributed models with better parallelism on existing computer hardware. DeepSpeed is optimized for low latency, high throughput training. It includes the Zero Redundancy Optimizer (ZeRO) for training models with 1 trillion or more parameters. Features include mixed precision training, single-GPU, multi-GPU, and multi-node training as well as custom model parallelism. The DeepSpeed source code is licensed under MIT License and available on GitHub.

Property	Value
dbo:abstract	DeepSpeed is an open source deep learning optimization library for PyTorch. The library is designed to reduce computing power and memory use and to train large distributed models with better parallelism on existing computer hardware. DeepSpeed is optimized for low latency, high throughput training. It includes the Zero Redundancy Optimizer (ZeRO) for training models with 1 trillion or more parameters. Features include mixed precision training, single-GPU, multi-GPU, and multi-node training as well as custom model parallelism. The DeepSpeed source code is licensed under MIT License and available on GitHub. The team claimed to achieve up to a 6.2x throughput improvement, 2.8x faster convergence, and 4.6x less communication. (en)
dbo:author	dbr:Microsoft_Research
dbo:developer	dbr:Microsoft
dbo:genre	dbr:Software_library
dbo:latestReleaseDate	2022-08-01 (xsd:date)
dbo:latestReleaseVersion	v0.7.0
dbo:license	dbr:MIT_License
dbo:programmingLanguage	dbr:Python_(programming_language) dbr:C++ dbr:CUDA
dbo:releaseDate	2020-05-18 (xsd:date)
dbo:thumbnail	wiki-commons:Special:FilePath/DeepSpeed_logo.svg?width=300
dbo:wikiPageExternalLink	https://arxiv.org/pdf/1910.02054.pdf%7Cyear=2019%7Carxiv=1910.02054 https://github.com/microsoft/DeepSpeed https://www.microsoft.com/en-us/research/blog/zero-deepspeed-new-system-optimizations-enable-training-models-with-over-100-billion-parameters/ https://www.microsoft.com/en-us/research/project/ai-at-scale/
dbo:wikiPageID	64396232 (xsd:integer)
dbo:wikiPageLength	4228 (xsd:nonNegativeInteger)
dbo:wikiPageRevisionID	1109254514 (xsd:integer)
dbo:wikiPageWikiLink	dbr:PyTorch dbr:Python_(programming_language) dbr:Deep_learning dbc:Free_and_open-source_software dbc:Microsoft_development_tools dbc:Microsoft_free_software dbc:Software_using_the_MIT_license dbr:Open_source dbr:GitHub dbr:Machine_learning dbr:Comparison_of_deep_learning_software dbr:Computer_hardware dbr:Parallel_computing dbr:C++ dbr:CUDA dbr:Distributed_computing dbr:TensorFlow dbc:C++_libraries dbc:Python_(programming_language)_libraries dbc:2020_software dbc:Microsoft_Research dbc:Deep_learning_software dbr:Memory_usage dbr:Microsoft dbr:Microsoft_Research dbr:MIT_License dbr:Software_library
dbp:author	dbr:Microsoft_Research
dbp:developer	dbr:Microsoft
dbp:genre	dbr:Software_library
dbp:latestReleaseDate	2022-08-01 (xsd:date)
dbp:latestReleaseVersion	v0.7.0 (en)
dbp:license	dbr:MIT_License
dbp:logo	DeepSpeed logo.svg (en)
dbp:name	DeepSpeed (en)
dbp:programmingLanguage	dbr:Python_(programming_language) dbr:C++ dbr:CUDA
dbp:released	2020-05-18 (xsd:date)
dbp:repo	https://github.com/microsoft/DeepSpeed
dbp:wikiPageUsesTemplate	dbt:Cite_journal dbt:Infobox_software dbt:Microsoft_FOSS dbt:Microsoft_Research dbt:Portal dbt:Reflist dbt:Short_description dbt:Start_date_and_age dbt:URL dbt:Microsoft-software-stub dbt:Deep_Learning_Software
dct:subject	dbc:Free_and_open-source_software dbc:Microsoft_development_tools dbc:Microsoft_free_software dbc:Software_using_the_MIT_license dbc:C++_libraries dbc:Python_(programming_language)_libraries dbc:2020_software dbc:Microsoft_Research dbc:Deep_learning_software
rdf:type	owl:Thing dbo:Software schema:CreativeWork dbo:Work wikidata:Q386724 wikidata:Q7397
rdfs:comment	DeepSpeed is an open source deep learning optimization library for PyTorch. The library is designed to reduce computing power and memory use and to train large distributed models with better parallelism on existing computer hardware. DeepSpeed is optimized for low latency, high throughput training. It includes the Zero Redundancy Optimizer (ZeRO) for training models with 1 trillion or more parameters. Features include mixed precision training, single-GPU, multi-GPU, and multi-node training as well as custom model parallelism. The DeepSpeed source code is licensed under MIT License and available on GitHub. (en)
rdfs:label	DeepSpeed (en)
owl:sameAs	wikidata:DeepSpeed https://global.dbpedia.org/id/D3wCv
prov:wasDerivedFrom	wikipedia-en:DeepSpeed?oldid=1109254514&ns=0
foaf:depiction	wiki-commons:Special:FilePath/DeepSpeed_logo.svg
foaf:isPrimaryTopicOf	wikipedia-en:DeepSpeed
foaf:name	DeepSpeed (en)
is dbo:wikiPageWikiLink of	dbr:PyTorch dbr:Timeline_of_artificial_intelligence dbr:Machine_learning dbr:Microsoft_and_open_source dbr:Link_prediction
is foaf:primaryTopic of	wikipedia-en:DeepSpeed