File size: 1,518 Bytes
3a96032 674121e 3a96032 674121e |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 |
---
language: Swedish
license: apache-2.0
---
# KB-BERT distilled base model (cased)
This model is a distilled version of [KB-BERT](https://huggingface.co/KB/bert-base-swedish-cased). It was distilled using Swedish data, the 2010-2015 portion of the [Swedish Culturomics Gigaword Corpus](https://spraakbanken.gu.se/en/resources/gigaword). The code for the distillation process can be found [here](https://github.com/AddedK/swedish-mbert-distillation/blob/main/azureML/pretrain_distillation.py). This was done as part of my Master's Thesis: *Task-agnostic knowledge distillation of mBERT to Swedish*.
## Model description
This is a 6-layer version of KB-BERT, having been distilled using the [LightMBERT](https://arxiv.org/abs/2103.06418) distillation method, but without freezing the embedding layer.
## Intended uses & limitations
You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to
be fine-tuned on a downstream task.
## Training data
The data used for distillation was the 2010-2015 portion of the [Swedish Culturomics Gigaword Corpus](https://spraakbanken.gu.se/en/resources/gigaword).
The tokenized data had a file size of approximately 7.4 GB.
## Evaluation results
When evaluated on the [SUCX 3.0 ](https://huggingface.co/datasets/KBLab/sucx3_ner) dataset, it achieved an average F1 score of 0.887 which is competitive with the score KB-BERT obtained, 0.894.
Additional results and comparisons are presented in my Master's Thesis
|