Views
No views yet
Table 2 - oBERT - SQuADv1 90% (in the upcoming updated version of the paper).Pruning method: oBERT upstream unstructured + sparse-transfer to downstream
Paper: https://arxiv.org/abs/2203.07259
Dataset: SQuADv1
Sparsity: 90%
Number of layers: 12(*)):| oBERT 90% | F1 | EM |
| ------------ | ----- | ----- |
| seed=42 | 88.55 | 81.48 |
| seed=3407 | 88.34 | 81.25 |
| seed=123 (*)| 88.64 | 81.57 |
| seed=12345 | 88.44 | 81.43 |
| ------------ | ----- | ----- |
| mean | 88.49 | 81.43 |
| stdev | 0.130 | 0.134 |1@article{kurtic2022optimal,
2 title={The Optimal BERT Surgeon: Scalable and Accurate Second-Order Pruning for Large Language Models},
3 author={Kurtic, Eldar and Campos, Daniel and Nguyen, Tuan and Frantar, Elias and Kurtz, Mark and Fineran, Benjamin and Goin, Michael and Alistarh, Dan},
4 journal={arXiv preprint arXiv:2203.07259},
5 year={2022}
6}