|
gptkbp:instance_of
|
gptkb:software_framework
|
|
gptkbp:aims_to
|
simplify distributed training
|
|
gptkbp:contributed_to
|
faster model training
|
|
gptkbp:developed_by
|
gptkb:Uber_Technologies
|
|
gptkbp:enables
|
multi-GPU training
multi-node training
|
|
gptkbp:has_community
|
active user community
|
|
gptkbp:has_documentation
|
available online
|
|
gptkbp:has_feature
|
dynamic scaling
data parallelism
model parallelism
gradient averaging
|
|
https://www.w3.org/2000/01/rdf-schema#label
|
Horovod
|
|
gptkbp:installation
|
pip install horovod
|
|
gptkbp:is_compatible_with
|
NVIDIAGP Us
CP Us
|
|
gptkbp:is_integrated_with
|
gptkb:lake
gptkb:fortification
gptkb:park
|
|
gptkbp:is_open_source
|
gptkb:theorem
|
|
gptkbp:is_optimized_for
|
large-scale training
|
|
gptkbp:is_part_of
|
gptkb:National_Park
deep learning frameworks
data science workflows
AI research projects
Tensor Flow ecosystem
Py Torch ecosystem
MX Net ecosystem
|
|
gptkbp:is_supported_by
|
gptkb:Microsoft_Azure
gptkb:DJ
gptkb:Google_Cloud
gptkb:AWS
gptkb:Intel
community contributions
|
|
gptkbp:is_used_by
|
industry leaders
research institutions
|
|
gptkbp:is_used_for
|
computer vision
natural language processing
reinforcement learning
image classification
|
|
gptkbp:is_used_in
|
gptkb:academic_research
Kaggle competitions
|
|
gptkbp:language
|
gptkb:Library
|
|
gptkbp:latest_version
|
0.23.0
|
|
gptkbp:performance
|
available on Git Hub
scales linearly with number of GP Us
|
|
gptkbp:provides
|
fault tolerance
easy integration with existing code
|
|
gptkbp:release_date
|
2017-09-01
|
|
gptkbp:released_in
|
gptkb:2017
|
|
gptkbp:repository
|
gptkb:archive
|
|
gptkbp:supports
|
gptkb:Graphics_Processing_Unit
gptkb:Py_Torch
MX Net
|
|
gptkbp:tutorials
|
available online
|
|
gptkbp:uses
|
gptkb:operating_system
Ring-All Reduce Algorithm
|
|
gptkbp:written_in
|
gptkb:C++
gptkb:Library
gptkb:CUDA
|
|
gptkbp:bfsParent
|
gptkb:DGX_A100
|
|
gptkbp:bfsLayer
|
5
|