Task Adapters

QNLI

Question Natural Language Inference is a version of SQuAD which has been converted to a binary classification task. The positive examples are (question, sentence) pairs which do contain the correct answer, and the negative examples are (question,sentence) from the same paragraph which do not contain the answer.

nli/qnli@ukp bert-base-uncased

1 version Architecture: houlsby

Adapter in Houlsby architecture trained on the QNLI task for 20 epochs with early stopping and a learning rate of 1e-4. See https://arxiv.org/pdf/2007.07779.pdf.

nli/qnli@ukp roberta-base

1 version Architecture: houlsby

QNLI adapter (with head) trained using the `run_glue.py` script with an extension that retains the best checkpoint (out of 15 epochs).

nli/qnli@ukp roberta-base

1 version Architecture: pfeiffer

QNLI adapter (with head) trained using the `run_glue.py` script with an extension that retains the best checkpoint (out of 15 epochs).

nli/qnli@ukp roberta-large

1 version Architecture: houlsby

QNLI adapter (with head) trained using the `run_glue.py` script with an extension that retains the best checkpoint (out of 15 epochs).

nli/qnli@ukp bert-base-uncased

1 version Architecture: pfeiffer

Adapter in Pfeiffer architecture trained on the QNLI task for 20 epochs with early stopping and a learning rate of 1e-4. See https://arxiv.org/pdf/2007.07779.pdf.

nli/qnli@ukp roberta-large

1 version Architecture: pfeiffer

QNLI adapter (with head) trained using the `run_glue.py` script with an extension that retains the best checkpoint (out of 10 epochs).