Inverse Square Root is a learning rate schedule 1 / $\sqrt{\max\left(n, k\right)}$ where $n$ is the current training iteration and $k$ is the number of warm-up steps. This sets a constant learning rate for the first $k$ steps, then exponentially decays the learning rate until pre-training is over.
Paper | Code | Results | Date | Stars |
---|
Task | Papers | Share |
---|---|---|
Language Modelling | 98 | 9.62% |
Question Answering | 66 | 6.48% |
Text Generation | 47 | 4.61% |
Sentence | 44 | 4.32% |
Translation | 32 | 3.14% |
Retrieval | 31 | 3.04% |
Machine Translation | 27 | 2.65% |
Natural Language Understanding | 22 | 2.16% |
Semantic Parsing | 19 | 1.86% |
Component | Type |
|
---|---|---|
🤖 No Components Found | You can add them if they exist; e.g. Mask R-CNN uses RoIAlign |