metaseq VS visual-chatgpt

This is a super important issue that affects the pace and breadth of iteration of AI almost as much as the raw hardware improvements do. The blog is fun but somewhat shallow and not technical or very surprising if you’ve worked with clusters of GPUs in any capacity over the years. (I liked the perspective of a former googler, but I’m not sure why past colleagues would recommend Jax over pytorch for LLMs outside of Google.) I hope this newco eventually releases a more technical report about their training adventures, like the PDF file here: https://github.com/facebookresearch/metaseq/tree/main/projec...

The number of people who can actually pre-train a true LLM is very small.

It remains a major feat with many tweaks and tricks. Case in point: the 114 pages of OPT175B logbook [1]

[1] https://github.com/facebookresearch/metaseq/blob/main/projec...

Reliability at scale: take a look at the OPT training log book for their 175B model run. It needed a lot of babysitting. In my experience, that scale of TPU training run requires a restart about once every 1-2 weeks—and they provide the middleware to monitor the health of the cluster and pick up on hardware failures.

I really appreciated this log of Facebook training a large language model of how troublesome AI development can be: https://github.com/facebookresearch/metaseq/tree/main/projects/OPT/chronicles

Stable Diffusion will run on any decent gaming GPU or a modern MacBook, meanwhile LLMs comparable to GPT-3/ChatGPT have had pretty insane memory requirements - e.g., <https://github.com/facebookresearch/metaseq/issues/146>

It seems so, check this log book from Meta: https://github.com/facebookresearch/metaseq/blob/main/projec...

For prompting, the GitHub repo I refer to the most is actually one from Microsoft: https://github.com/Microsoft/Visual-ChatGPT

Found relevant code at https://github.com/microsoft/visual-chatgpt + all code implementations here

Github: https://github.com/microsoft/visual-chatgpt/tree/main/LowCodeLLM will soon be available!

It's not so much the size of the LLMs, it's what people are able to do with them in novel ways, e.g. https://github.com/microsoft/visual-chatgpt/tree/main/TaskMatrix.AI

There is a python library langchain, which microsoft used to give it access to image editing tools. You can write your own tools for it very easily, I've been having fun with it. I gave it internet search tools and it worked great at using current events in the response.

Something like this: https://github.com/microsoft/visual-chatgpt/tree/main/TaskMatrix.AI could enable a far simpler AI to start making copies of itself.

Reminds me of VisualChatGPT (https://github.com/microsoft/visual-chatgpt), which also uses a LLM to decide what vision models to run.

The best Microsoft API or overall API? It really depends what your goal is. https://github.com/huggingface is great if you want to dive right in, but https://github.com/microsoft/visual-chatgpt might be good as well (not easy to get to work locally).