Skip to content
#

pretraining-model

Here are 5 public repositories matching this topic...

Language: All
Filter by language

Special-Fishstick introduces Windowed Masked Pretraining, an efficient BERT-style MLM approach where masking is applied globally while the encoder processes only random local windows. Achieves faster training and lower GPU memory with competitive language modeling performance.

  • Updated Aug 2, 2026
  • Jupyter Notebook

Improve this page

Add a description, image, and links to the pretraining-model topic page so that developers can more easily learn about it.

Curate this topic

Add this topic to your repo

To associate your repository with the pretraining-model topic, visit your repo's landing page and select "manage topics."

Learn more