Pith. sign in

Switch transformers: Scaling to trillion parameter models with simple and efficient sparsity

1 Pith paper cite this work. Polarity classification is still indexing.

1 Pith paper citing it

fields

cs.LG 1

years

2025 1

verdicts

CONDITIONAL 1

representative citing papers

Decoupled Relative Learning Rate Schedules

cs.LG · 2025-07-04 · conditional · novelty 6.0

Per-component relative learning-rate schedules speed up Transformer pretraining by up to 23% for MoE models, and the schedules tuned on a 34M model transfer to models 27x larger.

citing papers explorer

Showing 1 of 1 citing paper.

  • Decoupled Relative Learning Rate Schedules cs.LG · 2025-07-04 · conditional · none · ref 4

    Per-component relative learning-rate schedules speed up Transformer pretraining by up to 23% for MoE models, and the schedules tuned on a 34M model transfer to models 27x larger.