inari@piefed.zip to Piracy: ꜱᴀɪʟ ᴛʜᴇ ʜɪɢʜ ꜱᴇᴀꜱ@lemmy.dbzer0.comEnglish · 4 days agoPirate Face - a Pirate Bay for LLMspirateface.coexternal-linkmessage-square57linkfedilinkarrow-up1207 cross-posted to: hackernews@lemmy.bestiver.se
arrow-up1207external-linkPirate Face - a Pirate Bay for LLMspirateface.coinari@piefed.zip to Piracy: ꜱᴀɪʟ ᴛʜᴇ ʜɪɢʜ ꜱᴇᴀꜱ@lemmy.dbzer0.comEnglish · 4 days agomessage-square57linkfedilink cross-posted to: hackernews@lemmy.bestiver.se
minus-squareCethin@lemmy.ziplinkfedilinkEnglisharrow-up3·2 days agoMost consumers are not purchasing a GPU to run an LLM. They may choose to run one if they own a GPU. This does next to nothing for their demand. I don’t mean a subscription to Hugging Face. I mean a subscription to an LLM service. They don’t want you owning. They want you renting.
minus-squarenilloc@discuss.tchncs.delinkfedilinkEnglisharrow-up1·2 days agoI could see nvidia de-porting the hugging face models and making them harder to run on anything but their chips.
Most consumers are not purchasing a GPU to run an LLM. They may choose to run one if they own a GPU. This does next to nothing for their demand.
I don’t mean a subscription to Hugging Face. I mean a subscription to an LLM service. They don’t want you owning. They want you renting.
I could see nvidia de-porting the hugging face models and making them harder to run on anything but their chips.