metadata

tags:
  - generated_from_keras_callback
  - grammar error correction
model-index:
  - name: mt5-base-amhairc-gec-5
    results: []
language:
  - am
metrics:
  - bleu
base_model:
  - google/mt5-base
library_name: keras

mt5-base-amhairc-gec-5

This model was fine tuned on Amharic GEC dataset. It achieves the following results on the evaluation set: ~ (evaluating)

Model description

End-to-end Amharic grammar error correction, based on the mT5 model. Amharic is one of the most morphologically complex languages. Grammar corrections in this type of language are mostly made by applying morphology changes.

Intended uses & limitations

Losts of limitation, high accuracy was achived but with low precision. This is intended as proof of concept.

Training procedure

This model was finetuned for 5 epochs.

Training hyperparameters

The following hyperparameters were used during training:

optimizer: Adem
training_precision: float32

Training results

Framework versions

Transformers 4.35.2
TensorFlow 2.15.0
Datasets 2.14.5
Tokenizers 0.15.1