Task Adapters

MultiNLI

Multi-Genre Natural Language Inference is a large-scale, crowdsourced entailment classification task. Given a pair of sentences, the goal is to predict whether the second sentence is an entailment, contradiction, or neutral with respect to the first one.

Website

nli/multinli@kabirahuja2431 bert-base-multilingual-cased

2 versions Architecture: pfeiffer non-linearity: relu reduction factor: 2 Head:

Pfeiffer adapter stacked on top of language adapter for the NLI task. Trained on the English MultiNLI data for 5 epochs and a batch size of 64. Version 2 performs better for cross lingual transfer