Fine-grain compute communication execution for deep learning frameworks
One embodiment provides for a system to configure distributed training of a neural network.
The system includes memory to store a library to facilitate transmission of data during …
The system includes memory to store a library to facilitate transmission of data during …
Methods and apparatus for deep learning network execution pipeline on multi-processor platform
L Yang, A Yao - US Patent 11,461,105, 2022 - Google Patents
Methods and systems are disclosed using an execution pipeline on a multi-processor
platform for deep learning network execution. In one example, a network workload analyzer …
platform for deep learning network execution. In one example, a network workload analyzer …
Intelligent data transmission by network device agent
S Pasupuleti - US Patent 10,536,505, 2020 - Google Patents
In one aspect, a system for intelligent monitoring of a network device in a monitored
environment includes a processor; a memory; and one or more modules stored in the …
environment includes a processor; a memory; and one or more modules stored in the …
Matrix-factorization based gradient compression
Matrix factorization based gradient compression may be applied to an allreduce operation to
improve efficiency including the elimination of unnecessary meta data while maintaining …
improve efficiency including the elimination of unnecessary meta data while maintaining …
Dynamic network bandwidth in distributed deep learning training
Embodiments of a method are disclosed. The method includes performing distributed deep
learning training on a batch of training data. The method also includes determining training …
learning training on a batch of training data. The method also includes determining training …
Dynamic computation in decentralized distributed deep learning training
Embodiments of a method are disclosed. The method includes performing decentralized
distributed deep learning training on a batch of training data. Additionally, the method …
distributed deep learning training on a batch of training data. Additionally, the method …
Methods and apparatus for deep learning network execution pipeline on multi-processor platform
L Yang, A Yao - US Patent 11,868,782, 2024 - Google Patents
Methods and systems are disclosed using an execution pipeline on a multi-processor
platform for deep learning network execution. In one example, a network workload analyzer …
platform for deep learning network execution. In one example, a network workload analyzer …
Systems and methods for error recovery
B Pudipeddi, M Mesmakhosroshahi, J **… - US Patent …, 2023 - Google Patents
2021-12-16 Assigned to MICROSOFT TECHNOLOGY LICENSING, LLC reassignment
MICROSOFT TECHNOLOGY LICENSING, LLC ASSIGNMENT OF ASSIGNORS INTEREST …
MICROSOFT TECHNOLOGY LICENSING, LLC ASSIGNMENT OF ASSIGNORS INTEREST …
Dynamic computation rates for distributed deep learning
Embodiments of a method are disclosed. The method includes performing distributed deep
learning training on multiple batches of training data using corresponding learners …
learning training on multiple batches of training data using corresponding learners …
Systems and methods for error recovery
B Pudipeddi, M Mesmakhosroshahi, J **… - US Patent …, 2022 - Google Patents
2020-03-27 Assigned to MICROSOFT TECHNOLOGY LICENSING, LLC reassignment
MICROSOFT TECHNOLOGY LICENSING, LLC ASSIGNMENT OF ASSIGNORS INTEREST …
MICROSOFT TECHNOLOGY LICENSING, LLC ASSIGNMENT OF ASSIGNORS INTEREST …