scispace - formally typeset
M

Minjia Zhang

Researcher at Microsoft

Publications -  69
Citations -  1559

Minjia Zhang is an academic researcher from Microsoft. The author has contributed to research in topics: Computer science & Deep learning. The author has an hindex of 12, co-authored 45 publications receiving 737 citations. Previous affiliations of Minjia Zhang include Huazhong University of Science and Technology & Ohio State University.

Papers
More filters
Journal ArticleDOI

BLOOM: A 176B-Parameter Open-Access Multilingual Language Model

Teven Le Scao, +386 more
- 09 Nov 2022 - 
TL;DR: BLOOM as discussed by the authors is a decoder-only Transformer language model that was trained on the ROOTS corpus, a dataset comprising hundreds of sources in 46 natural and 13 programming languages (59 in total).
Proceedings ArticleDOI

Memcached Design on High Performance RDMA Capable Interconnects

Jithin Jose, +10 more
TL;DR: The design extends the existing open-source Memcached software and makes it RDMA capable and a detailed performance comparison of the Memcaches design is provided compared to unmodifiedmemcached using Sockets over RDMA and 10Gigabit Ethernet network with hardware-accelerated TCP/IP.
Posted Content

ZeRO-Offload: Democratizing Billion-Scale Model Training

Jie Ren, +7 more
- 18 Jan 2021 - 
TL;DR: ZeRO-Offload democratizes large-scale model training making it accessible to even data scientists with access to just a single GPU, and combines compute and memory efficiency with ease-of-use.
Proceedings Article

Learning Intrinsic Sparse Structures within Long Short-Term Memory

Wei Wen, +8 more
TL;DR: This work aims to learn structurally-sparse Long Short-Term Memory by reducing the sizes of basic structures within LSTM units, including input updates, gates, hidden states, cell states and outputs, by proposing Intrinsic Sparse Structures (ISS) in LSTMs.
Proceedings Article

DeepCPU: serving RNN-based deep learning models 10x faster

Minjia Zhang, +3 more
TL;DR: This work characterizes RNN performance and identifies low data reuse as a root cause, and develops novel techniques and an efficient search strategy to squeeze more data reuse out of this intrinsically challenging workload.