pkgx VS Llama-2-Onnx

I’m liking pkgx over asdf as it can activate project tooling upon cd’ing into a project folder.

https://pkgx.sh

I saw some alternatives being suggested and wanted to do the same (Also, so that I can look back at this item, through my comments :) ). Started using https://pkgx.sh/ lately. I know it has some baggage with tea.xyz and crypto, but it is also easy to get started with.

Secondly, our development environments must not drift, because then code may behave differently and a change could pass on our machine but fail in production. There are many tools for locking down environments, e.g nix, pkgx, asdf, containers, etc., and they all share the common goal of being able to lock down dependencies for an environment accurately and deterministically. And that needs to be enforced in our local workflow so we don't have to rely on CI environments for correctness. All developers must have environments that are effectively identical to what runs in CI (which itself should be representative of the production environment).

There are many ways this can be done (e.g nix, pkgx, asdf, containers, etc.), and we won’t get into which specific tools to use, because we'll instead cover the essential essence of preventing environment drift:

1. pkgx

Before diving in, it's a good idea to have a package manager handy, like tea. It'll handle your development environment and simplify your life!

Direct fast-track link to repo

checkout https://github.com/teaxyz/cli and https://github.com/teaxyz/pantry

Install is tea: sh <(curl https://tea.xyz) and

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