turmeric_root

> Isn't bragging the main thing people do on Social Media?

start following more interesting people

whisper doesn't seem to support diarization (identifying when the speaker changes), which is needed for subtitle formatting.

c'mon just write a function that takes in text and tells you whether or not it's true, how hard could it be

i think it'd be cool if you could make art and still own a house

> When users join a sub it's because they like the content and tacitly approve of the moderation.

I disagree, I see tons of people blaming admins for moderation actions and vice-versa.

clearly the arbitration panel ruled against Lindell because it's part of the deep state. /s

Windows reserves a certain percentage for VRAM for some reason. So I'd recommend Linux. Or find a way to disable the desktop/UI in Windows.

exactly! I gave a heartfelt letter to my shredder the other day and it simply destroyed it. issues like these are why AI alignment research is so critical.

Have you tried writing code on a smartphone?

this seems like it might be useful for arguing on HN

Worldcoin price skyrockets => cryptocurrency speculation returns => GPU shortage prevents further training of LLMs

Hell, what about Skynet?

The model weights were only shared by FB to people who applied for research access. Github repos containing links to the model weights have been taken down by FB.

I like using them for memeing

More VRAM => larger models. IME it is absolutely worth maxing out VRAM for the significant improvement in quality, especially with LLaMA (though even with a 4090, you won't be able to run the largest 65-billion parameter model even with 4-bit quantization).

That said, I recommend renting a cloud GPU for a few hours and trying the larger models on them before buying a GPU of your own, just to see if the models meet your requirements.

they're just microdosing it's ok

A lot of the 'look what I made with AI' images that get shared around also don't include the creator's workflow. There's usually lots of trial-and-error, manual painting/inpainting, multiple models involved etc. and explaining all that is a lot harder than just saying 'I used stable diffusion'.

ugh, that's so shitty. so many people in this space seem to be absurdly demanding and angry at devs, but one thing I've noticed is that every text AI project discord I've hung out in has this sleazy, obsessive 4chan /g/ vibe hiding somewhere in it.

> the "number B" stands for "number of billions" of parameters... trained on?

No, it's just the size of the network (i.e. number of learnable parameters). The 13/30/65B models were each trained on ~1.4 trillion tokens of training data (each token is around half a word).

'accuracy' and 'truth' are legacy 0.1X concepts, move fast and break things

yeah when getting DL up and running on AMD requires using a datacentre card then it's no wonder CUDA is more popular. AMD is enabling ROCm for commercial GPUs now but it's still a pain to get it up and running, because of the inertia that CUDA has.

if the AI is trained on LW then I think we'll be safe, just use the word 'woke' and it'll lose its shit and get stuck in an endless loop of telling you why it's not actually racist

though unless you've disabled sampling it will be difficult to determine how prompts affect the output, these could just be due to RNG

Yep.

I disagree with the linked post, most people use 'REST' to refer to JSON-over-HTTP now.

Yeah I spent a week or two getting excited playing with ChatGPT but then I got bored. I also bought a Quest 2 a while ago and sold it after a few months, so I guess the novelty just wears off quickly for me.

It seems to be about as good as gpt3-davinci. I've had it generate React components and write crappy poetry about arbitrary topics. Though as expected, it's not very good at instructional prompts since it's not tuned for instruction.

People are also working on adding extra samplers to FB's inference code, I think a repetition penalty sampler will significantly improve quality.

The 7B model is also fun to play with, I've had it generate Youtube transcriptions for fictional videos and it's generally on-topic.

So since making that comment I managed to get 65B running on 1 x A100 80GB using 8-bit quantization. Though I did need ~130GB of regular RAM on top of it.

> so does this mean you got it working on one GPU with an NVLink to a 2nd, or is it really running on all 4 A40s?

it's sharded across all 4 GPUs (as per the readme here: https://github.com/facebookresearch/llama). I'd wait a few weeks to a month for people to settle on a solution for running the model, people are just going to be throwing pytorch code at the wall and seeing what sticks right now.

the 7B model runs on a CUDA-compatible card with 16GB of VRAM (assuming your card has 16-bit float support).

I only got the 30b model running on a 4 x Nvidia A40 setup though.