oBERT-12-upstream-pretrained-dense

This model is obtained with The Optimal BERT Surgeon: Scalable and Accurate Second-Order Pruning for Large Language Models.

It corresponds to the pretrained dense model used as a teacher for upstream pruning runs, as described in the paper. The model can be finetuned on any downstream task, just like the standard bert-base-uncased model which is used as initialization for training of this model.

Sparse versions of this model:

• 90% sparse: neuralmagic/oBERT-12-upstream-pruned-unstructured-90
• 97% sparse: neuralmagic/oBERT-12-upstream-pruned-unstructured-97

Training objective: masked language modeling (MLM)
Paper: https://arxiv.org/abs/2203.07259
Dataset: BookCorpus and English Wikipedia
Sparsity: 0%
Number of layers: 12

Code: https://github.com/neuralmagic/sparseml/tree/main/research/optimal_BERT_surgeon_oBERT

If you find the model useful, please consider citing our work.

Citation info

@article{kurtic2022optimal,
  title={The Optimal BERT Surgeon: Scalable and Accurate Second-Order Pruning for Large Language Models},
  author={Kurtic, Eldar and Campos, Daniel and Nguyen, Tuan and Frantar, Elias and Kurtz, Mark and Fineran, Benjamin and Goin, Michael and Alistarh, Dan},
  journal={arXiv preprint arXiv:2203.07259},
  year={2022}
}