tart

Hardened VMs with no network devices and a serial console talking to it. There are so so so so many ways to do this. Anyone that built ISPs in the 90s can tell you this. Anyone that has built homelabs from scratch can. It is not that hard. It ain't easy. But it is not that hard. At. All. In fact, openai have https://github.com/openai/tart that can easily be adapted to more secure scenarios than whatever the f they are using atm.

> this won't speed up llama.cpp for everyone, just for users running it in this particular kind of Virtualization.framework VM. correct. these figures apply to llama.cpp inside the macOS guest configuration we tested. Lume is the VM frontend we used, while Apple's Virtualization.framework provides the virtual GPU. bare-metal llama.cpp is unaffected. > The fix here works around a problem where the VM was causing llama.cpp to select the wrong kernels. mostly, with one nuance: llama.cpp is selecting the correct kernels for the capability answers it receives. the stock guest reports an older Apple GPU family and a 32 KB threadgroup memory limit, so llama.cpp chooses slower kernels. Our process-scoped layer reports the tested Apple 9 and 64 KB values while allowing llama.cpp to select newer paths that the paravirtual GPU successfully execute t

What are your plans for tart licensing going forward? https://github.com/cirruslabs/tart/blob/main/LICENSE https://tart.run/licensing/

The title is a bit misleading then :) What’s the difference between this vs just using your lume CLI? Right now it feels like a worse interface to lume, but maybe I’m not getting a use case for this. Also, any thoughts on https://github.com/cirruslabs/tart? (alas, not open source)

the inability to log in with an Apple ID in a MacOS VM is a decision made by Apple. you can certainly sign without logging in, but I don't think you can submit to the store in a VM. you can see how they build tart here, including signing and entitlements and so on: https://github.com/cirruslabs/tart/blob/main/.cirrus.yml

https://github.com/cirruslabs/tart or brew install tart (this gets you the last open source version)

Tart is also very cool. It's not F/OSS but it is source-available and its license is liberal about non-commercial and small-scale commercial use: https://github.com/cirruslabs/tart The very cool thing about Tart is how it uses OCI for OS images, so you can use your existing image registry infrastructure to host and pull down OS images. Cilicon is a compatible F/OSS clone that currently only supports a subset of that functionality: https://github.com/traderepublic/Cilicon These are more CI-oriented but I like how that makes it easy to manage state with them. If you're focused on Linux VMs and maybe not on GUI stuff (although I'm sure you can make that work), Lima seems to be the go-to in the user 'community', as it were: https://github.com/lima-vm/lima If you use ARM guests on Apple Silicon, you should get good perf just like with stuff in

Boots in 5 minutes, minimum allocation of 24 hours, is "low stock" if you actually try to allocate one, and was recently "no stock". I'd be unsure if its actually available when you need it. They offered bare metal ARM some years ago and discontinued them (but I guess M1 hardware is a bit more reliable and/or popular enough [1]). Would be nice to see a Virtualization.Framework as a service, such as with Tart / Orchard [1], but perhaps there are performance and security issues when making those M1s multi-tenant. The license limits it to a maximum of two guest systems anyway, and only for the purposes of software development and testing [2]. [1] https://www.theregister.com/2020/04/21/scaleway_arm64_cloud_... [2] https://github.com/cirruslabs/tart/ [3] https://www.apple.com/legal/sla/docs/macOSVentura.pdf#page=2 (2B iii)

github.com/cirruslabs/tart is a command-line tool which makes this whole VM thing very easy on MacOS starting with Ventura.

github.com/cirruslabs/tart is what I use, because it is a command-line tool. also, death to oracle.

github.com/cirruslabs/tart does this for macos VMs. you push them to and pull them from OCI registries.