schizoidman@lemm.ee to Technology@lemmy.worldEnglish · 1 month agoDeepSeek's distilled new R1 AI model can run on a single GPU | TechCrunchtechcrunch.comexternal-linkmessage-square22linkfedilinkarrow-up1177arrow-down10 cross-posted to: technology@lemmy.zip
arrow-up1177arrow-down1external-linkDeepSeek's distilled new R1 AI model can run on a single GPU | TechCrunchtechcrunch.comschizoidman@lemm.ee to Technology@lemmy.worldEnglish · 1 month agomessage-square22linkfedilink cross-posted to: technology@lemmy.zip
minus-squarefogetaboutit@programming.devlinkfedilinkEnglisharrow-up4·1 month agoew probably still censored.
minus-squareT156@lemmy.worldlinkfedilinkEnglisharrow-up12·1 month agoThe censorship only exists on the version they host, which is fair enough. If they’re running it themselves in China, they can’t just break the law. If you run it yourself, the censorship isn’t there.
minus-squarejaschen@lemm.eecakelinkfedilinkEnglisharrow-up4·1 month agoUntrue, I downloaded the vanilla version and it’s hardcoded in.
minus-squareMonkderVierte@lemmy.mllinkfedilinkEnglisharrow-up2·edit-21 month agoYeah, i think the censoring in the LLM data itself would be pretty vulnerable to circumvention.
minus-squarejaschen@lemm.eecakelinkfedilinkEnglisharrow-up2·1 month agoThe self hosted model has hard coded censored content.
minus-squarefogetaboutit@programming.devlinkfedilinkEnglisharrow-up1·1 month agoif the model is censored… then what, retraining it? Or doing it from scratch like what open-r1 is doing?
ew probably still censored.
The censorship only exists on the version they host, which is fair enough. If they’re running it themselves in China, they can’t just break the law.
If you run it yourself, the censorship isn’t there.
Untrue, I downloaded the vanilla version and it’s hardcoded in.
Yeah, i think the censoring in the LLM data itself would be pretty vulnerable to circumvention.
You can self host it right??
The self hosted model has hard coded censored content.
if the model is censored… then what, retraining it? Or doing it from scratch like what open-r1 is doing?