Live data from Hacker News

Stretch iPhone to its limit: 2GiB Stable Diffusion model runs locally on device

liuliu.me

61–70 of 184 posts

Re: Stretch iPhone to its limit: 2GiB Stable Diffusion model runs locally on device

#61

This is cool and looks like a great way to drain my phone battery :) I just used the prompt "A person looking at their phone in amazement" and got a good picture. Beware that on startup the app downloads almost 2 gig of data.

It would be awesome if it just quickly took a picture from the front camera for that particular request and then just filter it to finish the "in the style of andy warhol"

Re: Stretch iPhone to its limit: 2GiB Stable Diffusion model runs locally on device

#62

the developer is about to have a MASSIVE hosting bill the download restarts from 0% if the app is sent to the background, as there does not seem to be a download manager. This is especially problematic for the large 1.5gb file.

I wonder why the file can’t be distributed inside the app on the App Store. That way downloads would be much more convenient.

appstore has app size limits.

Re: Stretch iPhone to its limit: 2GiB Stable Diffusion model runs locally on device

#63

Kinda ironic that this links to a site that is almost unreadable on an iPhone (14 pro max if you’re wondering).

I frequently see such comments about sites not loading fine on iPhone 13/14, while they continue to load just fine on for me on a 4 year old Android device (not pixel / Samsung). I wonder if it's the hardware or just the blockers that i use. Might be worth trying using blockers to see if it makes the general browsing experience better on Apple devices

It loads fine on an iPhone 14 Pro Max, it just has a very small font. Reader mode works great though.

Re: Stretch iPhone to its limit: 2GiB Stable Diffusion model runs locally on device

#64

the developer is about to have a MASSIVE hosting bill the download restarts from 0% if the app is sent to the background, as there does not seem to be a download manager. This is especially problematic for the large 1.5gb file.

I wonder why the file can’t be distributed inside the app on the App Store. That way downloads would be much more convenient.

Is P2P torrent type of sharing the load possible under AppStore guidelines (be they iOS or Android)? I've honestly never even thought about this being a thing, but with large shared data that doesn't change, why not?

Re: Stretch iPhone to its limit: 2GiB Stable Diffusion model runs locally on device

#65
post #6

This is cool and looks like a great way to drain my phone battery :) I just used the prompt "A person looking at their phone in amazement" and got a good picture. Beware that on startup the app downloads almost 2 gig of data.

It does warn you on startup about the download if you're not on wifi.

Good time to try 5G ultra capacity if you have it on an unlimited plan - will be faster than most people's wifi.

Re: Stretch iPhone to its limit: 2GiB Stable Diffusion model runs locally on device

#66

Earlier quoted context omitted.

I wonder why the file can’t be distributed inside the app on the App Store. That way downloads would be much more convenient.

appstore has app size limits.

There's also a way to host app resources in the App Store, usually used for levels in games and such:

https://developer.apple.com/library/archive/documentation/Fi...

Re: Stretch iPhone to its limit: 2GiB Stable Diffusion model runs locally on device

#68

Kinda ironic that this links to a site that is almost unreadable on an iPhone (14 pro max if you’re wondering).

The site loads fine but the font size is a little small for mobile. It reads fine if you rotate to landscape, though.

Re: Stretch iPhone to its limit: 2GiB Stable Diffusion model runs locally on device

#69
post #48

This is some impressive work. You might like to look at the work HuggingFace has been doing (on non-iOS versions). They can run it in under 1GB RAM: > If is also possible to chain it with attention slicing for minimal memory consumption, running it in as little as https://huggingface.co/docs/diffusers/optimization/fp16#offl...

CPU offloading doesn't work because Apple has shared memory arch already. The head slicing is similar to https://machinelearning.apple.com/research/neural-engine-tra... I think it is quite practical only if MPSGraph less mysterious about its allocation strategy. It is not the ideal way though. Ideally, FlashAttention / XFormer is the way to go.

Re: Stretch iPhone to its limit: 2GiB Stable Diffusion model runs locally on device

#70
Any tips on how you “ If the face is cropped, now I know how to use the inpainting model to fill it in. If the inpainting model doesn’t do its job, you can always use a paint brush to paint it over and do an image-to-image generation again focused in that area.” using the app?
Post reply on HN