Downloads · 30 days
16
20% of all-time downloads
Lambent/Arsenic-Shahrazad-12B-v4.4.2-mtp
Arsenic-Shahrazad-12B-v4.4.2-mtp is a machine learning model from Lambent. Use it for the machine learning task on the model card, and read the license before you ship it in a product. The card lists the license as cc-by-nc-4.0.
A version update meant to steer back towards v4.4's voice, using top-32 logits to strongly anchor towards the distribution while leaving the remaining ~2% probability mass available for MTP head co-adaptation.
Downloads · 30 days
16
20% of all-time downloads
All-time downloads
80
Public
Parameters
12.2B
24.5 GB on disk
Likes
1
Public
Click a slice to open those files.
.safetensors24.5 GB · 100%
From the Hugging Face model README
A version update meant to steer back towards v4.4's voice, using top-32 logits to strongly anchor towards the distribution while leaving the remaining ~2% probability mass available for MTP head co-adaptation.
Approximately matches the performance of v4.4.1's head and model.
Fork at https://github.com/lambentLogic/llama.cpp/tree/mtp-llama-mistral to actually utilize the head, upstream MTP keeps improving and I'll try to keep up with merging those in.