do you mean https://github.com/pollockjj/ComfyUI-MultiGPU? One GPU would do the computation, but others could pool in for VRAM expansion, right? (I've not used this node)

Nah, that won’t gain you much (if anything?) over just doing the layer swaps on RAM. You can put the text encoder on the second card but you can also just put it in your RAM without much for negatives.

I have 4x3090 (96GB) and 128GB DDR4 RAM, can I run unsloth on this machine and utilize all 4 GPUs?

There's also https://capibarazero.com/

RTX Pro 6000 Blackwell has 96GB VRAM.

It also costs 4x the entire Framework Desktop for just the card. If you're doing something professional that's probably worth it, but it's not a clear winner in the enthusiast space.

Thanks for the memory lane drive. I used to run this on FreeBSD desktop under fluxbox wm wayyy back, I think in 2004. So cool.

is this a wrapper to https://github.com/Tencent/Hunyuan3D-1/?

this is really cool and I'd love to use it, but it seems they only support workers written in Go if I'm not mistaken? My workers can be remote and not using Go, behind a NAT too. I want a them to periodically pull the queue, this way I do not need to worry about network topology. I guess I could simply interface with PG atomically via a simple API endpoint for the workers to connect, but I'd love to have the UI of riverqueue.

I gave them a year of subscriptions before cancelling recently, the devcontainer implementation in their Ultimate versions is laughably bad, bugs upon bugs and tickets where their support staff just bounces it up with "still no fix" messages and customers are finding workarounds, i.e. downgrading docker installs.

Remote SSH is terrible too, handles network latency spikes by repeating keystrokes. I remember spending an evening trying to fix something in the integrated shell and giving up, but sadly forgot what. I like what they do with Go though. Anyway, back to nvim here, not for me.

+1, the Remote SSH is horrible. Takes forever to connect and is extremely laggy once you have connected. Feels like they’re practically streaming video of the UI back to you instead of VSCode’s Remote SSH which feels indistinguishable from running locally.

Yeah I'm a big fan of the JetBrains IDEs but I tried it out a few days ago and couldn't even get it to stay connected. Gave up after reading a few recent forum posts about how much trouble people are still having with the feature (which does have a big purple "Beta" label in the IDE at least).

I'm hosting my blog on cloudflare pages, it's analytics show 80 or so uniques every day consistently even though I barely write there. Installed Umami - 0 visitors. None. Internet is just LLM crawlers hungry for content now?

We passed the tipping point where bot traffic outnumbered human traffic fifteen years ago. LLMs are an order of magnitude worse by most first-hand accounts, but it's just a continuation of a very long trend.

"Internet is just LLM crawlers hungry for content now?"

its been that way for a few years, real users using mobile app and access social media now

the percentage internet user who "surfing" on the web is dwindling and more likely diminish in near future

I see this too on my CF Pages-hosted blog.

Analytics only work if the agent runs JS. CF on the other hand counts file fetches, which can't be circumvented.

There's always a baseline of bot traffic.

ah, that explains it, I think. I expected them to sessionize the file transfers under one unique somehow still, even without JS.

You don't need a 512GB mac studio for this, TTS latency would be worse than 16GB 5080.