SuspciousCarrot78@lemmy.world to Privacy@lemmy.ml · edit-22 months agoI'm tired of LLM bullshitting. So I fixed it.codeberg.orgexternal-linkmessage-square136linkfedilinkarrow-up11arrow-down10file-textcross-posted to: [email protected][email protected]
arrow-up11arrow-down1external-linkI'm tired of LLM bullshitting. So I fixed it.codeberg.orgSuspciousCarrot78@lemmy.world to Privacy@lemmy.ml · edit-22 months agomessage-square136linkfedilinkfile-textcross-posted to: [email protected][email protected]
minus-squareDomi@lemmy.secnd.melinkfedilinkarrow-up0·5 months agoI have a Strix Halo machine with 128GB VRAM so I’m definitely going to give this a try with gpt-oss-120b this weekend.
minus-squareDomi@lemmy.secnd.melinkfedilinkarrow-up1·5 months agoOf course, self hosted with llama-swap and llama.cpp. :)
minus-squareDomi@lemmy.secnd.melinkfedilinkarrow-up1·5 months agogpt-oss is pretty much unusable without custom system prompt. Sycophancy turned to 11, bullet points everywhere and you get a summary for the summary of the summary.
minus-squarerecklessengagement@lemmy.worldlinkfedilinkarrow-up1·5 months agoStrix halo gang. Out of curiosity, what OS are you using?
I have a Strix Halo machine with 128GB VRAM so I’m definitely going to give this a try with gpt-oss-120b this weekend.
[deleted by user]
Of course, self hosted with llama-swap and llama.cpp. :)
[deleted by user]
gpt-oss is pretty much unusable without custom system prompt.
Sycophancy turned to 11, bullet points everywhere and you get a summary for the summary of the summary.
Strix halo gang. Out of curiosity, what OS are you using?
Fedora 43 with the Rawhide kernel.