test.mobilebert-uncased-squadv1
This model takes the same weights as neuralmagic/mobilebert-uncased-finetuned-squadv1 but is converted to OpenVINO IR format.
Eval results: (pad_to_max_length=True, max_length=384)
- torch model eval_exact_match: 83.9735
- torch model eval_f1: 90.9013
Framework versions
- Transformers 4.26.0
- Pytorch 1.13.1+cu116
- Datasets 2.8.0
- Tokenizers 0.13.2
- Optimum-intel 1.7.0
- NNCF 2.4.0
- Openvino-dev 2023.0.0.dev20230217
- Downloads last month
- 5
Dataset used to train yujiepan/test.mobilebert-uncased-squadv1
Evaluation results
- F1 on Squadv1validation set self-reported90.901
- EM on Squadv1validation set self-reported83.974