Best codeLlama model for query SQL generation
I have to choose the best model to generate sql code from natural language.
When I talk about the best model I mean the one that can best understand how to use sql operators like 'like' and that can be accurate even with rather long conditions.
Which one would you recommend between : codellama/CodeLlama-13b-Instruct-hf and codellama/CodeLlama-13b-hf (also considering 34b versions of parameters) ?
“Instruct” typically refers to providing guidance, directions, or explanations on how to perform a task you want AI to do whereas “chat” implies engaging in conversation or dialogue with someone.
If you are comfortable working with SQL commands (programmer, or such); for software or documentation, “instruct” is what you would use to explain how to do the tasks, and AI learns from you and perform tasks as you told it to, while you expect AI to already “know” these SQLs.
If you are not programmer and have basic knowledge how to do SQLs, then, chat is probably is one to use as chat can “share” with you the knowledge how to use SQL.