models and codes for baiduAI LIC 2021 MRC tasks, based on paddlenlp
Modified the baseline from paddlenlp.
Due to the poor push speed and the limitation of fold volume, dataset and model .pdparams are not pushed.
download.sh , downloading the dataset.
train.sh, train the model.
run_eval.sh, evaluation for the dev dataset.
prediction.sh, generate the demanding result.
| Rank | F1 | EM | in-domain(F1) | checklist(F1) | vocab(F1) | phrase(F1) | semantic-role(F1) | fault-tolerant(F1) | reasoning(F1) |
|---|---|---|---|---|---|---|---|---|---|
| 68 | 54.863 | 46.596 | 62.747 | 47.972 | 50.658 | 55.706 | 44.278 | 47.15 | 42.068 |
| Rank | F1 | EM | in-domain(F1) | checklist(F1) | vocab(F1) | phrase(F1) | semantic-role(F1) | fault-tolerant(F1) | reasoning(F1) |
|---|---|---|---|---|---|---|---|---|---|
| 85 | 58.89 | 48.774 | 66.659 | 52.6436 | 53.195 | 61.625 | 55.01 | 44.996 | 48.392 |
