hwers

8 ms·

31.
by hwers 4y ago
If you have even just 4gb stable diffusion will run fine if u go for 448x448 instead (basically the same quality).
32.
by hwers 4y ago
Here’s another one in gif form (right image) https://twitter.com/johnowhitaker/status/1565710033463156739 These seem really useful to get intuition about im2im
33.
Stable Diffusion UI for Rapid Prototyping(github.com)
3 points by hwers 4y ago|0 comments
34.
by hwers 4y ago
We could always do im2tex via just clip embedding the image. The idea that you could hide/sell prompts is silly. (Having human interpretable im2tex is cool tho.)
35.
by hwers 4y ago
This message brought to you by replicate.com
36.
by hwers 4y ago
I have 4 gb and it takes about 9 seconds for me :) (Tho at 448x448 but there’s no real difference in quality.)
37.
by hwers 4y ago
Whats wasm32 (vs just normal wasm)?
38.
by hwers 4y ago
whats the difference between float16 and bfloat16?
39.
by hwers 4y ago
Same. Things are getting so noisy and hard to find that even if good howtos exist somewhere on the internet, finding it can waste hours vs just word searching your own documents.
40.
by hwers 4y ago
That’s because $20 is still humiliatingly low for the workload
41.
by hwers 4y ago
Yeah stylegan is rarely this good at super diverse data like this
42.
by hwers 4y ago
Lol imagine if someone said this as a prediction when photoshop was invented.
43.
by hwers 4y ago
> It puzzles me when people say that small companies can’t retrain every day because they don’t have FAANG-style budgets. It costs a few dollars, at best, to retrain many xgboost or scikit-learn models. Most models are not large language
44.
by hwers 4y ago
What are you using it for if can share? I’ve thought about training some of these and releasing the weights but I’ve never found a reason they’d really be useful personally so it never really happened
45.
Denoising Diffusion at 1 second per output image(developer.nvidia.com)
2 points by hwers 4y ago|0 comments
46.
by hwers 4y ago
Anyone know what SIMD means at the bottom layer is here? I know what SIMD is I mean in the context of the preprocessor (and being the worst offender apparently).
47.
by hwers 4y ago
He mentions open source and education which I’m really excited about. There’s been a big lack of accessible from-first-principles resources on this (plenty of abstracted libraries though). Now is a great time to build up excitement among am
48.
by hwers 4y ago
Surely 1bn could have afforded a lot more and better marketing..
49.
by hwers 4y ago
So embarrassing if he ends up being forced to own it and everyone knows he tried to back out.
50.
by hwers 4y ago
This is such a terrible look for Elon. I’ve really liked him up until now but this just feels like playing with our emotions. Honestly starting to question how much hope there is for mars to happen at all.
51.
by hwers 4y ago
I don’t even think it blocks the channel long term from your recommendations. I’m pretty sure they’ve tried to sneak in an old channel I’ve explicitly said not to recommend again in my recommendations via whatever multi armed bandit with no
52.
by hwers 4y ago
I guess an interesting way to translate this technique to text-to-image would be to get an image from a database that matches the text query (via CLIP) and feed that + noise into a diffusion model that only does a few denoising iterations (
53.
by hwers 4y ago
New product lines are always on the horizon
54.
by hwers 4y ago
Love that analogy at the end, very well said
55.
by hwers 4y ago
It’s silly that you’re being downvoted because this is absolutely true. The idea that it would go to zero is mad, it will without a doubt go back up to similar numbers again. This is just a little gift
56.
by hwers 4y ago
This whole “what matters is the founder” idea really discourages me. It’s such a VC perspective. The VC has a flock of people they pick from to invest in and if your traits doesn’t fit the necessary requirements you shouldn’t start a startu
57.
Oimo.io(oimo.io)
1 points by hwers 4y ago|0 comments
58.
by hwers 4y ago
Well my point was that you could extract it into a wasm blob without onnx (library independent). See tensor4 as linked in the other thread.
59.
by hwers 4y ago
It's just a form of stress coping. It's entirely delusional and I think it's very important we not allow this type of thinking to be a normal form of thought (as uncomfortable as it is).
60.
by hwers 4y ago
Yup. In torch you can use torchscript and in jax you can 'jit' a function and both turn your NN function into a fairly easily compliable form which could then be translated to wasm. Here's a cool project that does something l

More ›