EDiT trains LLMs with local SGD plus model sharding, adding a pseudo-gradient penalty to stabilize loss; it reports modest quality gains and higher throughput, though a proof of convergence contains an error.
Local sgd optimizes overparameterized neural networks in polynomial time
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.DC 1years
2024 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
EDiT: A Local-SGD-Based Efficient Distributed Training Method for Large Language Models
EDiT trains LLMs with local SGD plus model sharding, adding a pseudo-gradient penalty to stabilize loss; it reports modest quality gains and higher throughput, though a proof of convergence contains an error.