ML Reading #32: Modular LLM Fine-tuning

​LlamaFactory: Unified Efficient Fine-Tuning of 100+ Language Models

A Survey on LoRA of Large Language Models

Tinker: a training API for researchers and developers

We’ll explore emerging scaffolding around tools like LoRA and the new ecosystems that promise to make customized AI both more powerful and more accessible - like Thinking Machine’s Tinker API. Join us at Mox to discuss what “efficient adaptation” truly means as these methods rapidly shift the boundaries of model capability.

​Some questions to get us started:

• ​Are we moving toward a world of countless specialized micro-models?

• ​What might an agentic RL system look like?

​

🔎Analyzed Papers

​​​Discussion at 20:00, (optional) quiet reading from 19:00.