Howard: Developers should distribute merged adapters rather than merged models
Jeremy Howard · Answer.ai & AI Magic with Jeremy Howard · Aug 17, 2024 · at 46:26
Jeremy Howard (Answer.ai / fast.ai) explains why distributing quantized base models with merged LoRA adapters is superior to distributing full merged FP16 model checkpoints.
0:00 / 0:57exact quote · 57.7s
▶ Watch the full episode on YouTube → 720p mp4 · rendered on demand · StarZero watermark
“To explain, it's not that you shouldn't merge models, it's that you shouldn't be distributing a merged model. You should distribute it a merged adapter. 99% of the time. And actually often, one of the best things happening in the model merging world is actually that often merging adapters works better. The point is, Sean, that, that once you've got your new model, If you distribute it as an adapter that sits on top of a quantized model that somebody's already downloaded, then it's a much smaller download for them, and also the inference should be much faster, because you're not having to transfer FB-sixteen weights from FB, from HPM memory at all, or ever load them off disk. You know, all the main weights are quantized, and the only floating point weights are in the adapters. So that should make both inference and fine tuning faster.”
quote is from the automated transcript, cleaned for reading: filler sounds and stutters are removed, nothing is rephrased. names can be misheard (the analysis reads context, assessments check outside sources). how →
More from Jeremy Howard
Howard: Alec Radford Built OpenAI's GPT After Reading ULMFiT
“I organized a chat for both of us with Kate Metz in the New York Times, and Kate Metz answered, sorry, and Alec answered this question for Kate, and Kate just like, so how did, you know, GPT come about? And he said, well, I was pretty sure that pre-training on…”
Howard: Meta 'blew it' on Code Llama due to catastrophic forgetting
“So Code Llama was a, I think it was like a five hundred billion token fine-tuning of Llama II using code. And also prose about code that Meta did. And honestly, they kind of blew it. Because Code Llama is good at coding, but it's bad at everything else.”
Howard: TensorFlow 2 was a failure that Google avoided internally
“Then in the end, you know, Google didn't follow through, which is fair enough, like, asking everybody to, you know, learn a new programming language is going to be tough, but, like, it was very obvious, very, very obvious at that time that TensorFlow II was go…”
Howard: JAX was a grassroots Google reaction against TensorFlow 2
“But I mean, in the meantime, I will say, you know, Google now does have a backup plan. You know, they have JAX, which was never a strategy. It was just a bunch of people who also recognized TensorFlow two as shit, and they just decided to build something else.”
Howard: RAG is an inefficient hack compared to fine-tuning
“RAG is like such a inefficient hack, really, isn't it? It's like, You know, segment up my data in some somewhat arbitrary way, embed it, ask questions about that, you know, hope that my embedding, you know, model embeds questions in the same embedding space as…”
Howard: Answer.AI runs fully in-house stack without AWS or Google Cloud
“This group of, which has averaged about 10 to 12 people, currently nine, I think, have built a pretty Transformational and complex piece of software, which we can do a quick demo of later if you're interested. Using a complete web application development platf…”