Get completions from a model fine-tuned with large pool of text

[tardy obsidian]

currently able to fine tune by giving Q&A dataset. when a text is passed, each token is getting embedding. and i am not able to get completion for my question from the embeddings