GBA: a tuning-free approach to switch between synchronous and asynchronous training for recommendation models
W Su, Y Zhang, Y Cai, K Ren, P Wang… - Advances in …, 2022 - proceedings.neurips.cc
High-concurrency asynchronous training upon parameter server (PS) architecture and high-
performance synchronous training upon all-reduce (AR) architecture are the most commonly …
performance synchronous training upon all-reduce (AR) architecture are the most commonly …
BachLedger: Orchestrating Parallel Execution with Dynamic Dependency Detection and Seamless Scheduling
Y Yang, G Shang, G Qi, Z Ma, Y Liu… - 2024 IEEE 30th …, 2024 - ieeexplore.ieee.org
Blockchain technology inherently necessitates redundant computation to achieve
consensus among untrusted parties because of its fundamental threat model. This …
consensus among untrusted parties because of its fundamental threat model. This …
[PDF][PDF] Load-centric data shuffling with a patch-based repartitioning algorithm exploiting the data placement and distribution
E Kassela - 2024 - dspace.lib.ntua.gr
This diploma thesis aims to the development of a novel data repartitioning algorithm that can
be used to process unordered skewed data in distributed environments. In workloads …
be used to process unordered skewed data in distributed environments. In workloads …