CalamitousFelicitousness/Llama-3.05-Nemotron-Tenyxchat-Storybreaker-70B-FP8-Dynamic

This repo contains the copy of the original quantized to FP8. Original: Envoid/Llama-3.05-Nemotron-Tenyxchat-Storybreaker-70B

Llama-3.05-Nemotron-Tenyxchat-Storybreaker-70B

is a 40/60 SLERP Merge of Envoid/Llama-3-TenyxChat-DaybreakStorywriter-70B onto nvidia/Llama-3.1-Nemotron-70B-Instruct-HF utilizing the following config:

models:
  - model: ./Envoid_Llama-3-TenyxChat-DaybreakStorywriter-70B
  - model: ./nvidia_Llama-3.1-Nemotron-70B-Instruct-HF
merge_method: slerp
base_model: ./nvidia_Llama-3.1-Nemotron-70B-Instruct-HF
parameters:
  t:
    - value: 0.4
dtype: bfloat16

Caution: As is always the case with SLERP merges there may be edge cases inwhich certain unintended model behaviors emerge. So always use with caution.

The 'sloppiness' of Nemotron seems to be somewhat reigned in (but still exists) while maintaining its personable assistant personality and safety (In assistant mode it will still prompt you with a warning before producing sensitive content).

Overall it provides a solid option for RP and creative writing while still functioning as an assistant model, if desired. If used to continue a roleplay it will generally follow the ongoing cadence of the conversation.

It utilizes the Llama 3 prompt format.