akhooli_-_gpt2-small-arabic-8bits

Quantization made by Richard Erkhov.

Github

Discord

Request more models

gpt2-small-arabic - bnb 8bits

• Model creator: https://huggingface.co/akhooli/
• Original model: https://huggingface.co/akhooli/gpt2-small-arabic/

Original model description:

language: "ar" datasets:

• Arabic Wikipedia metrics:
• none

---

GPT2-Small-Arabic

Model description

GPT2 model from Arabic Wikipedia dataset based on gpt2-small (using Fastai2).

Intended uses & limitations

How to use

An example is provided in this colab notebook. Both text and poetry (fine-tuned model) generation are included.

Limitations and bias

GPT2-small-arabic (trained on Arabic Wikipedia) has several limitations in terms of coverage (Arabic Wikipeedia quality, no diacritics) and training performance. Use as demonstration or proof of concepts but not as production code.

Training data

This pretrained model used the Arabic Wikipedia dump (around 900 MB).

Training procedure

Training was done using Fastai2 library on Kaggle, using free GPU.

Eval results

Final perplexity reached was 72.19, loss: 4.28, accuracy: 0.307

BibTeX entry and citation info

@inproceedings{Abed Khooli,
  year={2020}
}