Nothing's training yet.

A training takes something you want a model to learn - your conversations, your notes, how you write - and teaches it to a model you pick. You don't need to know how. Start one and it'll tell you what it's doing, in words, the whole way.

Ran one before? Start from its config:
1 What should it learn? ?
Test it on: ?
Add a dataset
Examples: one per line (JSONL) - conversations, ShareGPT, instruction/output, prompt/completion or plain text. Questions: one per line, as plain text or JSON, with the answer you expect if you have one.
2 Start from which model? ?
or any model on Hugging Face:
3 Where should it train?
4 How hard should it study?
Advanced (the numbers behind it)

Leave these alone and the level you picked sets them. Change one and yours wins.

5 What should come out? ?
Pick a dataset to see what's in it.
Asking this box what it has…