Llama-2-Onnx VS towhee

System: Here's some docs, answer concisely in a sentence.

YMMV on cost still, depends on cloud vendor, and my intuition & viewpoint agrees with yours, GPT-3.5 is priced low enough that there isn't a case where it makes sense to use another model.

It strikes me now that _very_ likely and not just our intuition: OpenAI's $/GPU hour is likely <= any other vendor's.

The next big step will come from formalizing the stuff rolling around the local LLM community, for months now it's either been one-off $X.c stunts that run on desktop, and the vast majority of the _actual_ usage and progress is coming from porn-y stuff, like all nascent tech.

Microsoft has LLaMa-2 ONNX available on GitHub[1]. There's budding but very small projects in different languages to wrap ONNX. Once there's a genuine cross-platform[2] ONNX wrapper that makes running LLaMa-2 easy, there will be a step change. It'll be "free"[3] to run your fine-tuned model that does as well as GPT-4 .

It's not clear to me exactly when this will occur. It's "difficult" now, but only because the _actual usage_ in the local LLM community doesn't have a reason to invest in ONNX, and it's extremely intimidating to figure out how exactly to get LLaMa-2 running in ONNX. Microsoft kinda threw it up on GitHub and moved on, the sample code even still needs a PyTorch model. I see at least one very small company on HuggingFace that _may_ have figured out full ONNX.

[1] https://github.com/microsoft/Llama-2-Onnx

As another commenter noted, Milvus is overkill and a "bit much" if you're learning/playing.

A good intro to the field with progression towards a full Milvus implementation could be starting with towhee[0] (which is also supported by Milvus).

towhee has an example to do exactly what you want with CLIP[1].

[0] - https://towhee.io/

[1] - https://github.com/towhee-io/examples/tree/main/image/text_i...

The description of this is kind of confusing but I think the easiest way to understand it is that it is a data processing pipeline of sorts. Take unstructured data and apply transformation and computation. A similar project to this is Towhee (https://github.com/towhee-io/towhee). This project tries to simplify unstructured data processing and provides pretrained models and pipelines from their hub.

Milvus also has incredible flexibility when it comes to choosing an indexing strategy, and we also have a library specifically meant to help vectorize a variety of data called Towhee (https://github.com/towhee-io/towhee).

Benchmarking the models with towhee is as simple as:

The project's homepage is https://github.com/towhee-io/towhee, and you can find more about towhee by going through the documents.

I made a tutorial for building an image search engine with python. The code example is as simple as 10 lines of code, using Towhee and Milvus To put images into the search engine:

Traditionally, I've done this through PyTorch by adding a hook, but this requires knowledge of the model itself (i.e. model arch and layer names). I found https://github.com/Hironsan/awesome-embedding-models but it didn't provide many CV-focused open-source projects. There's also https://github.com/towhee-io/towhee which is great but more targeted towards application development.

You can check the result from the tutorial.