Fine-Tune Arabic Model

[Omar Samir]

I have created a dataset with almost 200 million words. So there are about 20 million examples to train the model on if each image contains 10 words. Is it enough to get better results? 

under consideration, we have fine-tuned a model using 20 thousand examples and it did worse than the pre-trained model.