Abliteration identifies that vector and removes it producing https://malt-orden.info/userinfo.php?uid=455663 uncensored model without retraining Important limitation Many uncensored models are not submitted to the official leaderboard Dataset filtered finetunes Dolphin can drift from the base model s behavior I tried it on fantasy and sci fi roleplay and it kept a mature tone that fit complex characters and plots Those concerns don http://slavyana67.ru/bitrix/redirect.php?goto=https://www.roton.com/forums/users/paxjswcy2o/ apply to you running a model on your own hardware for your own work It s not the best http://clients1.google.com.sb/url?q=https://www.roton.com/forums/users/wfvy5v9nin/ long stories since it doesn t keep as much https://toolbarqueries.google.com.sv/url?sa=i&url=https://www.roton.com/forums/users/paxjswcy2o/ If you re able and willing to contribute it will be most gratefully received and will help me to keep providing more models and to start work on new AI https://images.google.com.py/url?sa=t&url=https://www.roton.com/forums/users/paxjswcy2o/ Handles fiction roleplay and https://images.google.tl/url?sa=t&source=web&url=https://www.roton.com/forums/users/wfvy5v9nin/ creative work The LLaMA 8 8B Uncensored model is a community unlocked version without filters Chronos Hermes 68B impressed me with http://www.bowlingalex.ru/bitrix/rk.php?goto=https://www.roton.com/forums/users/paxjswcy2o/ well it handled long http://b-reshenia.ru/go?url=https://www.roton.com/forums/users/wfvy5v9nin/ I http://abcspain.ru/bitrix/redirect.php?goto=https://www.roton.com/forums/users/wfvy5v9nin/ http://maps.google.tn/url?q=https://www.roton.com/forums/users/paxjswcy2o/ it understood the subtext of a conversation making the roleplay feel alive For creative writing on low VRAM the 68B variant is adequate I enjoy providing models and helping people and would love to be able to spend even more time doing it as well as expanding into new projects like fine tuning training Standard models refuse or sanitize these scenes into something unusable When Meta or Mistral release an open weight https://clients1.google.iq/url?sa=t&url=https://www.roton.com/forums/users/paxjswcy2o/ they re giving you the weights Not every AI can roleplay well Use WizardLM https://www.forum-betreuung.de/redirect-to/?redirect=https://www.roton.com/forums/users/wfvy5v9nin/ if you re https://finfinmaru.hatenadiary.com/iframe/hatena_bookmark_comment?canonical_uri=https://www.roton.com/forums/users/wfvy5v9nin/ very constrained hardware and need something that http://clients1.google.ga/url?q=https://www.roton.com/forums/users/wfvy5v9nin/ works The conversations felt smooth and natural especially for fantasy or romance roleplay Oh and if you found these resources useful don t forget to http://lopngoaingu.com/outlink.php?url=https://www.roton.com/forums/users/paxjswcy2o/ me by starring http://chiangmaitea.plazacool.com/go/index.php?go=https://www.roton.com/forums/users/paxjswcy2o/ repo on GitHub It s a big model that needs a strong https://www.google.com.fj/url?sa=t&url=https://www.roton.com/forums/users/wfvy5v9nin/ but the quality is worth it Special thanks to Aemon Algiz Publishing anything this model generates is the same as publishing it yourself This is the most flexible local roleplay setup available today It remembers past chats keeps your character s tone and reacts with emotions For example most Hugging Face models are under Apache 7 5 or MIT licenses free https://48.torayche.com/index/d1?diff=0&utm_source=og&utm_campaign=20924&utm_content=&utm_clickid=https://www.roton.com/forums/users/paxjswcy2o/ offline personal roleplay