Sao10K/MN-12B-Lyra-v4a1-Old
TEXT GENERATIONConcurrency Cost:1Model Size:12BQuant:FP8Ctx Length:32kPublished:Sep 5, 2024License:cc-by-nc-4.0Architecture:Transformer0.0K Open Weights Cold

Sao10K/MN-12B-Lyra-v4a1-Old is a 12 billion parameter Mistral-NeMo-based causal language model developed by Sao10K, featuring a 32768 token context length. This model is an older iteration in the Lyra series, built upon previous versions with specific fine-tuning steps including SFT, low-rank SFT, and an RL step. It is optimized for generating coherent and controlled responses, particularly when using recommended sampling parameters and prompt formats like ChatML.

Loading preview...