oBERT-12-upstream-pruned-unstructured-90-finetuned-mnli

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

It corresponds to the model presented in the Table 2 - oBERT - MNLI 90%.

Pruning method: oBERT upstream unstructured + sparse-transfer to downstream
Paper: https://arxiv.org/abs/2203.07259
Dataset: MNLI
Sparsity: 90%
Number of layers: 12

The dev-set performance reported in the paper is averaged over three seeds, and we release the best model (marked with (*)):

| oBERT 90%    | m-acc | mm-acc|
| ------------ | ----- | ----- |
| seed=42   (*)| 82.40 | 83.40 |
| seed=3407    | 82.15 | 83.41 |
| seed=54321   | 82.32 | 83.38 |
| ------------ | ----- | ----- |
| mean         | 82.29 | 83.40 |
| stdev        | 0.127 | 0.015 |

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}
}