Abdullah Al Asif
Abdullah Al Asif
@kausikMR Yes, the .ipynb should work for Tamil as well. You don't have to change much except how the data is processed, which would depend on your dataset's structure (file...
@kausikMR A good test would be to check if the tokenizer can convert Tamil text to token IDs. If it can do that, you should be able to fine-tune it...
@kausikMR I heard `lora` doesn't perform well for Orpheus language adoption. Maybe you can try the full training.
@kausikMR, I could only train the `lora` version.