Learning Rate Schedules
Inverse Square Root Schedule
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.
Papers
| Paper | Code | Results | Date | Stars |
|---|
Tasks
| Task | Papers | Share |
|---|---|---|
| Language Modelling | 78 | 9.30% |
| Question Answering | 61 | 7.27% |
| Text Generation | 45 | 5.36% |
| Retrieval | 26 | 3.10% |
| Natural Language Understanding | 23 | 2.74% |
| Machine Translation | 22 | 2.62% |
| Semantic Parsing | 19 | 2.26% |
| Abstractive Text Summarization | 18 | 2.15% |
| Natural Language Inference | 16 | 1.91% |
Usage Over Time
Components
| Component | Type |
|
|---|---|---|
| 🤖 No Components Found | You can add them if they exist; e.g. Mask R-CNN uses RoIAlign |
