fasthmtl fine-tuned model

[RonanKMcGovern]

Hi folks, is there an LLM that has been fine-tuned with fasthtml data?

If one were to do such a fine-tuning, what would be your brief recommendations for data preparation?

• DPO or SFT or both?
• Just use the examples in this repo? Somehow expand them to create more data?
• Blend in a small percentage of some good code data from elsewhere?

Thanks! and thanks for having made this.