I am utilizing the tsdocode/text-to-sql model in order to build an app for obviously converting text to sql. I was able to complete the process of encoding, decoding, submitting data, and training it, but I don’t think I’m training it properly. My outputs from the “fine-tuned” model are treating the database columns as if their also plaintext where it should be using the columns names it was trained on. This problem has lead me to believe that I evidently not formatted the data properly. My csv consists of one column consisting of a plaintext question and the second column is its sql query counterpart. I simply plugged this into load_dataset() and went from there. How do I load this dataset/ specify this data so that the model knows the first column is supposed to become the second column.
Any and all advice would be greatly appreciated.
You can share you jupyter notebook by colab so others can get better understanding of you question.