Awni Hannun
Awni Hannun
@smsharma this looks great!! I pushed a few mostly cosmetic changes to your branch. It runs and it works well :). One thing I wanted to run by you: As...
Also I moved the example from `flow` to `real_nvp` to be a bit more descriptive / help with discovery
It's a great idea, we'd love to extend the LoRA example to more models. You can see how we [convert TinyLlama here](https://github.com/ml-explore/mlx-examples/blob/main/llama/convert.py#L51-L111)
There are a couple things missing in the LoRA example to work with TinyLlama (namely RoPE without traditional).
I got it working yesterday by using the conversion script in the `llama` example with a couple of small changes in the `lora` example. I think it makes sense to...
Wow awesome!! > Would there be interest in this? Definitely. Is it possible to fetch and build a version of the library with CMake rather than include the whole thing...
> What's the best path to land this? Do you prefer one big PR with as many features as possible or a small one? Your check list looks reasonable to...
> I think this may already be useful for some people, so perhaps we can merge it as is and do any improvements later. What do you think? I will...
Sent a few diffs, here. It looks in great shape to me. One thing I find inconsistent with the rest of our save routines is that the open mode is...
I have a couple more questions about the functionality: Right now we don't support loading any quantized formats natively, but we default to load `fp32`. I'm wondering if that's the...