DALLE2-pytorch VS metaseq

I was showing people Dalle2 last year and telling them how much of an impact an open source solution was going to have on, well, everything to do with art and design. (At the time Stable Diffusion had not released, not even the leak, and all hopes was on https://github.com/lucidrains/DALLE2-pytorch)

Stable diffusion is the model, MJ will use a variant and DALL-E is the old version (we have our own implementation from our distinguished fellow Lucidrains here: https://github.com/lucidrains/DALLE2-pytorch)

Sadly I don't think so. From what I understand the architecture is fixed to 1024x1024 pictures.

The code is open source, the model is not I believe. https://github.com/lucidrains/DALLE2-pytorch

Phil Wang has been fairly prolific at creating open source implementations of these text to image models. For example, here is the dalle-2 repo https://github.com/lucidrains/DALLE2-pytorch

There's already an open-source implementation of DALL-E 2 (https://github.com/lucidrains/DALLE2-pytorch) and a pretrained model for it should be released within this year.

Also true for Google's Imagen, which should be even better than DALLE-2 (and faster) https://github.com/lucidrains/imagen-pytorch.

This is possible because the original research papers behind both DALLE-2 and Imagen were publicly released.

The paper describing the model is public and has been implemented here, but that's not the hard part. The model likely requires months of compute and dozens of gigabytes of VRAM to train and run, likely costing several hundred thousand dollars.

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...