File size: 2,681 Bytes
3530341 |
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 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 |
---
language: ar
datasets:
- oscar
- wikipedia
tags:
- ar
- masked-lm
---
# Arabic-ALBERT Xlarge
Arabic edition of ALBERT Xlarge pretrained language model
_If you use any of these models in your work, please cite this work as:_
```
@software{ali_safaya_2020_4718724,
author = {Ali Safaya},
title = {Arabic-ALBERT},
month = aug,
year = 2020,
publisher = {Zenodo},
version = {1.0.0},
doi = {10.5281/zenodo.4718724},
url = {https://doi.org/10.5281/zenodo.4718724}
}
```
## Pretraining data
The models were pretrained on ~4.4 Billion words:
- Arabic version of [OSCAR](https://oscar-corpus.com/) (unshuffled version of the corpus) - filtered from [Common Crawl](http://commoncrawl.org/)
- Recent dump of Arabic [Wikipedia](https://dumps.wikimedia.org/backup-index.html)
__Notes on training data:__
- Our final version of corpus contains some non-Arabic words inlines, which we did not remove from sentences since that would affect some tasks like NER.
- Although non-Arabic characters were lowered as a preprocessing step, since Arabic characters do not have upper or lower case, there is no cased and uncased version of the model.
- The corpus and vocabulary set are not restricted to Modern Standard Arabic, they contain some dialectical Arabic too.
## Pretraining details
- These models were trained using Google ALBERT's github [repository](https://github.com/google-research/albert) on a single TPU v3-8 provided for free from [TFRC](https://www.tensorflow.org/tfrc).
- Our pretraining procedure follows training settings of bert with some changes: trained for 7M training steps with batchsize of 64, instead of 125K with batchsize of 4096.
## Models
| | albert-base | albert-large | albert-xlarge |
|:---:|:---:|:---:|:---:|
| Hidden Layers | 12 | 24 | 24 |
| Attention heads | 12 | 16 | 32 |
| Hidden size | 768 | 1024 | 2048 |
## Results
For further details on the models performance or any other queries, please refer to [Arabic-ALBERT](https://github.com/KUIS-AI-Lab/Arabic-ALBERT/)
## How to use
You can use these models by installing `torch` or `tensorflow` and Huggingface library `transformers`. And you can use it directly by initializing it like this:
```python
from transformers import AutoTokenizer, AutoModel
# loading the tokenizer
tokenizer = AutoTokenizer.from_pretrained("kuisailab/albert-xlarge-arabic")
# loading the model
model = AutoModelForMaskedLM.from_pretrained("kuisailab/albert-xlarge-arabic")
```
## Acknowledgement
Thanks to Google for providing free TPU for the training process and for Huggingface for hosting these models on their servers 😊
|