From 823f15af4c40d88c5ffeddc2711a3db8364ed39d Mon Sep 17 00:00:00 2001 From: Moritz Laurer Date: Sun, 5 Dec 2021 17:47:23 +0000 Subject: [PATCH] Update README.md --- README.md | 8 +++++--- 1 file changed, 5 insertions(+), 3 deletions(-) diff --git a/README.md b/README.md index 03868a8..e8a77e7 100644 --- a/README.md +++ b/README.md @@ -33,7 +33,7 @@ widget: --- # Multilingual mDeBERTa-v3-base-mnli-xnli ## Model description -This multilingual model can perform NLI on 100+ languages. It was pre-trained by Microsoft on the [CC100 multilingual dataset](https://huggingface.co/datasets/cc100). It was then fine-tuned on the [XNLI dataset](https://huggingface.co/datasets/xnli), which contains hypothesis-premise pairs from 15 languages as well as the English [MNLI dataset](https://huggingface.co/datasets/multi_nli). +This multilingual model can perform natural language inference (NLI) on 100 languages and is therefore also suitable for multilingual zero-shot classification. The underlying model was pre-trained by Microsoft on the [CC100 multilingual dataset](https://huggingface.co/datasets/cc100). It was then fine-tuned on the [XNLI dataset](https://huggingface.co/datasets/xnli), which contains hypothesis-premise pairs from 15 languages, as well as the English [MNLI dataset](https://huggingface.co/datasets/multi_nli). As of December 2021, mDeBERTa-base is the best performing multilingual transformer (base) model, introduced by Microsoft in [this paper](https://arxiv.org/pdf/2111.09543.pdf). @@ -56,7 +56,7 @@ print(prediction) ``` ### Training data -This model was trained on the XNLI development dataset and the MNLI train dataset. The XNLI development set consists of 5010 professionally translated texts for each of 15 languages (see [this paper](https://arxiv.org/pdf/1809.05053.pdf)). Note that the XNLI contains a training set of 15 machine translated versions of the MNLI dataset for 15 languages, but due to quality issues with these machine translations, this model was only trained on the professional translations from the XNLI development set and the original English MNLI training set (392 702 texts). Not using machine translated texts can avoid overfitting the model to the 15 languages and avoid catastrophic forgetting of the other 85 languages mDeBERTa was pre-trained on. +This model was trained on the XNLI development dataset and the MNLI train dataset. The XNLI development set consists of 5010 professionally translated texts for each of 15 languages (see [this paper](https://arxiv.org/pdf/1809.05053.pdf)). Note that the XNLI contains a training set of 15 machine translated versions of the MNLI dataset for 15 languages, but due to quality issues with these machine translations, this model was only trained on the professional translations from the XNLI development set and the original English MNLI training set (392 702 texts). Not using machine translated texts can avoid overfitting the model to the 15 languages; avoids catastrophic forgetting of the other 85 languages mDeBERTa was pre-trained on; and significantly reduces training costs. ### Training procedure DeBERTa-v3-base-mnli was trained using the Hugging Face trainer with the following hyperparameters. @@ -71,7 +71,9 @@ training_args = TrainingArguments( ) ``` ### Eval results -The model was evaluated on the XNLI test set. Note that if other multilingual models on the model hub claim performance of around 90% on languages other than English, the authors have most likely made a mistake during testing since non of the latest papers shows a multilingual average performance of more than a few points above 80% on XNLI (see [here](https://arxiv.org/pdf/2111.09543.pdf) or [here](https://arxiv.org/pdf/1911.02116.pdf)). +The model was evaluated on the XNLI test set on 15 languages. Note that multilingual NLI models are capable of classifying NLI texts without receiving NLI training data (cross-lingual transfer). This means that the model is also able of doing NLI on the other 85 languages mDeBERTa was training on, but performance is most likely lower than for those languages available in XNLI. + +Also note that if other multilingual models on the model hub claim performance of around 90% on languages other than English, the authors have most likely made a mistake during testing since non of the latest papers shows a multilingual average performance of more than a few points above 80% on XNLI (see [here](https://arxiv.org/pdf/2111.09543.pdf) or [here](https://arxiv.org/pdf/1911.02116.pdf)). average | ar | bg | de | el | en | es | fr | hi | ru | sw | th | tr | ur | vu | zh ---------|----------|---------|----------|----------|----------|----------|----------|----------|----------|----------|----------|----------|----------|----------|----------