Orthogonalized models
Hey @mradermacher
Thank you for all the hard work!
I found a few models that might be of interest:
These first 2 are pending a merge request by me to fix the tokenizer_config.json to match the official llama-3.
Llama-3-8B-Instruct-abliterated: https://huggingface.co/failspy/Llama-3-8B-Instruct-abliterated
Llama-3-8B-Instruct-abliterated-v2: https://huggingface.co/cognitivecomputations/Llama-3-8B-Instruct-abliterated-v2
llama-3-70B-Instruct-abliterated: https://huggingface.co/failspy/llama-3-70B-Instruct-abliterated
mistral-orthogonalized: https://huggingface.co/cosmicvalor/mistral-orthogonalized
kappa-3-phi-abliterated: https://huggingface.co/failspy/kappa-3-phi-abliterated
:)
I've added the last three to the queue - just drop me a note when the first two are updated and I add them, too.
Thank you! And will do :)
@mradermacher they have been updated/merged :)
Wow, so quick. Let's quantize them!
Thank you so much!