Task Adapters

SQuAD 2.0

SQuAD2.0 combines the 100,000 questions in SQuAD1.1 with over 50,000 unanswerable questions written adversarially by crowdworkers to look similar to answerable ones. To do well on SQuAD2.0, systems must not only answer questions when possible, but also determine when no answer is supported by the paragraph and abstain from answering.

qa/squad2@ukp roberta-base

1 version Architecture: pfeiffer Head:

Adapter for roberta-base in Pfeiffer architecture trained on the SQuAD 2.0 dataset for 15 epochs with early stopping and a learning rate of 1e-4.

qa/squad2@ukp bert-base-uncased

1 version Architecture: pfeiffer Head:

Adapter for bert-base-uncased in Pfeiffer architecture trained on the SQuAD 2.0 dataset for 15 epochs with early stopping and a learning rate of 1e-4.

qa/squad2@ukp bert-base-uncased

1 version Architecture: houlsby Head:

Adapter for bert-base-uncased in Pfeiffer architecture trained on the SQuAD 2.0 dataset for 15 epochs with early stopping and a learning rate of 1e-4.

qa/squad2@ukp roberta-base

1 version Architecture: houlsby Head:

Adapter for roberta-base in Houlsby architecture trained on the SQuAD 2.0 dataset for 15 epochs with early stopping and a learning rate of 1e-4.