Fine Tuning over custom queries
Hi, I was wondering if it was possible to fine tune this model over my custom database schema queries. If yes, can i use the usual fine tuning LoRA procedure for Codellama by just replacing base codellama with sqlcoder-34b-alpha.
Thanks.
Is it possible to do something like this https://github.com/ragntune/code-llama-finetune/blob/main/fine-tune-code-llama.ipynb with custom queries and defog/sqlcoder-34b-alpha
@dobby-remastered Any updates on this from your end?
@srivatsan32 Yes, we can fine tune the model in a similar way https://github.com/ragntune/code-llama-finetune/blob/main/fine-tune-code-llama.ipynb have done with custom dataset to increase accuracy on specific database schema, as in-house database schema could not be inferred correctly by model for multiple columns causing ambiguity. Fine tuning by creating a specific dataset helped generalize the model for propriety database schema.