How do you train it ?
Hello,
thank you for sharing. I want to make myself a finetuned mistral model with custom dataset. Please tell me how did you trained yours if you like. How do you make a finetune lora model joined with the original mistral model into a single finetuned model. I saw other guys that make a distinct finetune model than the original one ..then load both. I'm interested to know how to make 1 single model joined from original mistral + finetuned model , like yours. Thnak you for help!
@Moneymaker2023 Hello,
Check this project https://github.com/OpenAccess-AI-Collective/axolotl
is a great finetuning tool , all what you need in one place. It support many types of FT, lora/qlora, with deepspeed and many other features, like merge adapter to base model.