roberta-base-amharic

This model has the same architecture as xlm-roberta-base and was pretrained from scratch using the Amharic subsets of the oscar, mc4, and amharic-sentences-corpus datasets, on a total of 290 Million tokens. The tokenizer was trained from scratch on the same text corpus, and had a vocabulary size of 32k.

The model was trained for 22 hours on an A100 40GB GPU.

It achieves the following results on the evaluation set:

• Loss: 2.09
• Perplexity: 8.08

This model has 110 Million parameters and is currently the best Amharic encoder model, beating the 2.5x larger 279 Million parameter xlm-roberta-base multilingual model on Amharic Sentiment Classification and Named Entity Recognition tasks.

How to use

You can use this model directly with a pipeline for masked language modeling:

>>> from transformers import pipeline
>>> unmasker = pipeline('fill-mask', model='rasyosef/roberta-base-amharic')
>>> unmasker("ከሀገራቸው ከኢትዮጵያ ከወጡ ግማሽ ምዕተ <mask> ተቆጥሯል።")

[{'score': 0.40162667632102966,
  'token': 137,
  'token_str': 'ዓመት',
  'sequence': 'ከሀገራቸው ከኢትዮጵያ ከወጡ ግማሽ ምዕተ ዓመት ተቆጥሯል።'},
 {'score': 0.24096301198005676,
  'token': 346,
  'token_str': 'አመት',
  'sequence': 'ከሀገራቸው ከኢትዮጵያ ከወጡ ግማሽ ምዕተ አመት ተቆጥሯል።'},
 {'score': 0.15971705317497253,
  'token': 217,
  'token_str': 'ዓመታት',
  'sequence': 'ከሀገራቸው ከኢትዮጵያ ከወጡ ግማሽ ምዕተ ዓመታት ተቆጥሯል።'},
 {'score': 0.13074122369289398,
  'token': 733,
  'token_str': 'አመታት',
  'sequence': 'ከሀገራቸው ከኢትዮጵያ ከወጡ ግማሽ ምዕተ አመታት ተቆጥሯል።'},
 {'score': 0.03847867250442505,
  'token': 194,
  'token_str': 'ዘመን',
  'sequence': 'ከሀገራቸው ከኢትዮጵያ ከወጡ ግማሽ ምዕተ ዘመን ተቆጥሯል።'}]

Finetuning

This model was finetuned and evaluated on the following Amharic NLP tasks

• Sentiment Classification
  • Dataset: amharic-sentiment
  • Code: https://github.com/rasyosef/amharic-sentiment-classification
• Named Entity Recognition
  • Dataset: amharic-named-entity-recognition
  • Code: https://github.com/rasyosef/amharic-named-entity-recognition

Finetuned Model Performance

The reported F1 scores are macro averages.