🇵🇭 phi3 finetune+genai

[trajepl]

Describe your changes

Existing genai example only cover the case of export model from official huggingface hub. This PR shows how to use model builder to optimize the finetune lora weights.

Note that: genai currently only works for LORA finetuned model. Qlora will use nf4 or fp4 which leads several issues in genai:

1. nf4/fp4 will force the model on cuda when loading. This lead runtime error when convert weights from tensor to numpy (need fix from genai).
2. weights shape dismatch error. (investigating...)

Checklist before requesting a review

• Add unit tests for this change.
• Make sure all tests can pass.
• Update documents if necessary.
• Lint and apply fixes to your code by running lintrunner -a
• Is this a user-facing change? If yes, give a description of this change to be included in the release notes.
• Is this PR including examples changes? If yes, please remember to update example documentation in a follow-up PR.

(Optional) Issue link

[devang-ml]

Please update Phi3 example readme.