Fine-tune a smaller BERT model for ALDi estimation

Use the following code-snippet:

```
# Set the ID of the GPU device
CUDA_ID="0"

# Fine-tune the model
SEED="42"
MODEL_NAME="SMALL_ARABIC_BERT_MODEL" # TODO

CUDA_VISIBLE_DEVICES="$CUDA_ID" python finetune_BERT_models.py train --train data/AOC/train.tsv --dev data/AOC/dev.tsv -model_name "$MODEL_NAME" -o Sentence_ALDi -s "$SEED"
```