It really depends on how you quantize the model and the K/V cache as well. This is a useful calculator. https://smcleod.net/vram-estimator/ I can comfortably fit most 32b models quantized to 4-bit (usually KVM or IQ4XS) on my 3090’s 24 GB of VRAM with a reasonable context size. If you’re going to be needing a much larger context window to input large documents etc then you’d need to go smaller with the model size (14b, 27b etc) or get a multi GPU set up or something with unified memory and a lot of ram (like the Mac Minis others are mentioning).
FrankLaskey
joined 5 years ago
It would be more interesting to see this with a cost of living figure for each state as well.
Is it possible to use StreetComplete on iOS?
I think we can all agree that modifications to these models which remove censorship and propaganda on behalf of one particular country or party is valuable for the sake of accuracy and impartiality, but reading some of the example responses for the new model I honestly find myself wondering if they haven’t gone a bit further than that by replacing some of the old non-responses and positive portrayals of China and the CPC with a highly critical perspective typified by western governments which are hostile to China (in particular the US). Even the name of the model certainly doesn’t make it sound like neutrality and accuracy is their primary aim here.
Yeah I use voyager pretty much exclusively on my iPhone so maybe I should request a feature like that there? Seems like it would be something that many people would appreciate. Not sure why I end up seeing posts with -10, -15 votes.. Those are generally trash haha
Thanks for the info. I don’t know a ton about them but I’m honestly massively impressed at the talent of the Proton devs. The fact that they have made most games run as well and some games run better on Unix operating systems through a translation layer than on Windows (the OS they were designed for) is ridiculously impressive. And this just shows they aren’t resting on their laurels but are being proactive in preventing issues before they happen which is immensely commendable and impressive.
Thanks for the context. I figured I must be missing something since performance increases like this for proton would be huge news and likely not possible.
Anyone have more context on this? These are some pretty massive increases if the games in the table are in any way representative of all games.
By having China create a more efficient and capable government and open source it so you can run it locally?
Oh and I typically get 16-20 tok/s running a 32b model on Ollama using Open WebUI. Also I have experienced issues with 4-bit quantization for the K/V cache on some models myself so just FYI