FlexGen VS autocomplete

LLaMA, Pythia, RWKV, Flan-T5 (self-hosted), FlexGen

> With no real knowledge of LLM and only recently started to understand what LLM terms mean, such as 'model, inference, LLM model, intruction set, fine tuning' whatelse do you think is required to make a took like yours?

This was mee a few weeks ago. I got interested in all this when FlexGen (https://github.com/FMInference/FlexGen) was announced, which allowed to run inference using OPT model on consumer hardware. I'm an avid user of Stable Diffusion, and I wanted to see if I can have an SD equivalent of ChatGPT.

Not understanding the details of hyperparameters or terminology, I basically asked ChatGPT to explain to me what these things are:

   Explain to someone who is a software engineer with limited knowledge of ML terms or linear algebra, what is "feed forward" and "self-attention" in the context of ML and large language models. Provide examples when possible.

Having contributed to the Fig autocomplete specs, I find this sad. The Amazon product Fig was built into basically works as replacement, which is good. Still, the core value of this product are the open-source autocomplete specs: https://github.com/withfig/autocomplete. What's going to happen to that? It looks like they are still using it in the Amazon product. It should definitely be possible for an open-source re-implementation of the Fig UI to use those specs. There is a lot of knowledge encoded in there!

8. Fig

Fig is awesome but with signups blocked[1] for 2+mo already it's also as good as dead ¯\_(ツ)_/¯

* [1]: https://github.com/withfig/autocomplete/issues/2068

https://github.com/withfig/autocomplete is it this?

How is this different than https://fig.io/?

AWS CloudWatch Evidently The worst. No comment. AWS seems to perpetually lack a good DX for developers. It appears that they don't recognize or continually undervalue the importance of roles other than engineers, such as Product Managers or Designers. Very disappointing. However, AWS has recently acquired Fig, so looks like they're now pursuing an acquisition strategy instead. Let's see how it turns it out, and let's hope they don't ruin Fig, since it's such an useful tool.

slightly tangential, but where do people get awesome landing pages like linear(https://fig.io/. has similar landing page) etc. Do they build them in-house or buy templates somewhere? Many of the recently launched YC companies have awesome landing pages. eg. https://automorphic.ai/,

I love this product, have contributed several times to it, and I'm a little torn. One thing I am thinking about now, is that the completion specs are MIT-licensed, and it should be possible to use them to re-implement a basic open-source version of the autocompletion product... https://github.com/withfig/autocomplete