Overview
Language model: deepset/roberta-base-squad2-distilled
Language: English
Training data: SQuAD 2.0 training set
Eval data: SQuAD 2.0 dev set
Infrastructure: 4x V100 GPU
Published: Dec 8th, 2021
Details
- haystack's distillation feature was used for training. deepset/roberta-large-squad2 was used as the teacher model.
Hyperparameters
batch_size = 80
n_epochs = 4
max_seq_len = 384
learning_rate = 3e-5
lr_schedule = LinearWarmup
embeds_dropout_prob = 0.1
temperature = 1.5
distillation_loss_weight = 0.75
Performance
"exact": 79.8366040596311
"f1": 83.916407079888
Authors
Timo Möller: timo.moeller@deepset.ai
Julian Risch: julian.risch@deepset.ai
Malte Pietsch: malte.pietsch@deepset.ai
Michel Bartels: michel.bartels@deepset.ai
About us
![](https://raw.githubusercontent.com/deepset-ai/.github/main/deepset-logo-colored.png)
![](https://raw.githubusercontent.com/deepset-ai/.github/main/haystack-logo-colored.png)
deepset is the company behind the open-source NLP framework Haystack which is designed to help you build production ready NLP systems that use: Question answering, summarization, ranking etc.
Some of our other work:
- Distilled roberta-base-squad2 (aka "tinyroberta-squad2")
- German BERT (aka "bert-base-german-cased")
- GermanQuAD and GermanDPR datasets and models (aka "gelectra-base-germanquad", "gbert-base-germandpr")
Get in touch and join the Haystack community
For more info on Haystack, visit our GitHub repo and Documentation.
We also have a Discord community open to everyone!
Twitter | LinkedIn | Discord | GitHub Discussions | Website
By the way: we're hiring!
- Downloads last month
- 50,505
Dataset used to train deepset/roberta-base-squad2-distilled
Spaces using deepset/roberta-base-squad2-distilled 7
Evaluation results
- Exact Match on squad_v2validation set verified80.859
- F1 on squad_v2validation set verified84.010
- Exact Match on squadvalidation set self-reported86.225
- F1 on squadvalidation set self-reported92.483
- Exact Match on adversarial_qavalidation set self-reported29.900
- F1 on adversarial_qavalidation set self-reported41.183
- Exact Match on squad_adversarialvalidation set self-reported79.071
- F1 on squad_adversarialvalidation set self-reported84.472
- Exact Match on squadshifts amazontest set self-reported70.733
- F1 on squadshifts amazontest set self-reported83.958