glizzyguzzler@lemmy.blahaj.zone to 196@lemmy.blahaj.zoneEnglish · 10 months agoFF rulelemmy.blahaj.zoneimagemessage-square12fedilinkarrow-up1403arrow-down10file-text
arrow-up1403arrow-down1imageFF rulelemmy.blahaj.zoneglizzyguzzler@lemmy.blahaj.zone to 196@lemmy.blahaj.zoneEnglish · 10 months agomessage-square12fedilinkfile-text
minus-squareAmaltheamannenlinkfedilinkarrow-up15·10 months agoI upgraded to 64gb, which is surprisingly cheap nowadays. Plus it allows you to run quantised LLMs locally.
minus-squareben@lemmy.ziplinkfedilinkEnglisharrow-up4·10 months agoOnly downside is the memory training times on reboots.
minus-squareGladaed@feddit.delinkfedilinkarrow-up4·10 months agoIsn’t it supposed to Do that only once?
minus-squareTwanHE@lemmy.worldlinkfedilinkarrow-up3·10 months agoYou can set manual timings and disable the training
minus-squareBloodSlut@lemmy.worldlinkfedilinkarrow-up1·10 months agoshouldnt there be ways around this? like taking a snapshot of memory and saving it to disk
I upgraded to 64gb, which is surprisingly cheap nowadays. Plus it allows you to run quantised LLMs locally.
Only downside is the memory training times on reboots.
Isn’t it supposed to Do that only once?
You can set manual timings and disable the training
shouldnt there be ways around this?
like taking a snapshot of memory and saving it to disk