How to load finetuned checkpoints using LoRA and DoRA

[abhijeetkumar1]

Hi, I’m working on zero-shot forecasting and looking to load fine-tuned checkpoints using LoRA and DoRA. Could you provide guidance on the recommended approach for doing this? Specifically:

What’s the best way to load LoRA/DoRA fine-tuned adapters for inference?
Are there any example scripts or configurations you recommend?
Do I need to merge the adapters with the base model before inference, or can I directly apply them dynamically?
Any documentation or pointers would be greatly appreciated. Thanks in advance for your help!

Best,
Abhijeet Kumar