>NVIDIA to buy hugging face>Includes ownership of llama.cppSay goodbye to innovation in open models and local models and genning and getting off of NVIDIA hardware. NVIDIA will also be able to see trends occurring in realtime before any competitor can
Is Hugging Face the only one of its kind? Is there really no alternative?
>>109675584Based nvidiaGODS
>>109675584Get fucked. We're going to the fucking moon. T. NVIDIA INVESTORSAILING TO THE STARS MOTHER FUCKER
Another Novelai victory.
>>109675611Its not simply about hosting model files anon. Anyone can do that. https://youtu.be/8_FjjgbQpKs?si=DKuJqxXKSSl4_AhQ>Access to data and trends before competitorsNvidia can control and steer the market in their favour. Nvidia will do what's good for maintain their monopoly. Did you enjoy the already jewish pricing pre-AI? How about now? Do you want to have the option of getting off nvidia hardware for whatever current thing is locally?
Should AMD just drop out of the AI game at this point?
>>109675754That guy also mentioned that Huawei bypassed the need for cuda at all.
Isnt HF mostly known for distributing assload of data to people for free
I voted for this
>>109675754That'll result in what is essentially a technological fork between ai hardware and non-ai hardware. Sounds pretty great if it actually happens.
>>109675754No, because they can still fill the gap in the market NVIDIA has abandoned. Everyone wants local AI but nobody wants to pay out their ass for it.AMD can do a good enough job of that even if I wish their consumer cards had more VRAM for it.
>>109676122It won't ever been "non AI hardware" as they are already used, hence future GPUs will as well. I don't think they will move to NPUs either for a lot of things for future models, they'd need the high compute of GPUs. And then "retro" AI will run on them too (today's models)
>>109676133Intel had very cheap 16GB cards and recently has a 32GB card. Which would have been a lot cheaper if not for the corpos kvetching over accessible and cheap memory chips. Though some models are also compute bound. I've been running H3 minimax on my 16GB 4080 and only 32GB DDR4 system ram. I have a paging file on my NM790 nvme which is enough to keep everything fed, the GPU is continually maxed out bar a few seconds in the middle of a gen where there's a dip in GPU and spike in the NVME out of a 10 min gen. High vram is unlikely to make this faster.
>>109676352>>109676431gues what, is not our problem.
>>109676461>No contextNo one knows what this is supposed to mean
suck a thousand balls if you support any kind of corpo-monopoly
>>109675584How is it allowed? Can't regulator see they are allowing monopoly to be created. This will bring absolutely nothing good.
>>109677106They're already getting away with round trip circular financing
>>109676030NOT ANYMORE.
>>109677106AI companies are military assets so governments won't let regulations cross them.
