All credit goes to the original model authors, this is just a double back slerp merge (config in repo) i did while commiting many various dataset war-crimes. I mean... ShareGPT Formaxxing.
Other Notes: Text completion preset in repo - Instruct format works with either Mistral or ChatML. If using ST with chatML add ["<|im_end|"] to custom stopping strings to prevent chatml eos tokens from leaking into chat.
Downloads last month
22
Safetensors
Model size
12.2B params
Tensor type
BF16
·
Inference API
Unable to determine this model's library. Check the
docs
.
Model tree for Nitral-AI/Lyra_Gutenbergs-Twilight_Magnum-12B