I just published 3 things:
1. A LoRA that adds Romanian diacritics to texts that don't have them: huggingface.co/sqreept/ro_dia… (includes an example of using it)
2. The dataset used to build the above LoRA: huggingface.co/datasets/sqree…
3. The Colab used to fine-tune the above LoRA: colab.research.google.com/drive/1GMg9fS3…
This is what I call open source in Machine Learning: model weights, dataset and code. Anything less is not open source.
