BruhzWater/Apocrypha-L3.3-70b-0.4a
BruhzWater/Apocrypha-L3.3-70b-0.4a is a 70 billion parameter Llama-3.3-based merged language model, developed by BruhzWater, with a 32768 token context length. This model is specifically optimized for storytelling and creative writing tasks. It integrates components from several Llama-3.3-70B models, including 'Llama-3-70B-Instruct-Storywriter' and 'Wayfarer-Large-70B-Llama-3.3', to enhance narrative generation and response completion.
Loading preview...
Apocrypha-L3.3-70b-0.4a Overview
Apocrypha-L3.3-70b-0.4a is a 70 billion parameter language model built on the Llama-3.3 architecture, designed primarily for storytelling and creative writing. This model is a merge of several specialized Llama-3.3-70B variants, utilizing the SCE merge method with BruhzWater/Eden-L3.3-70b-0.4a as its base.
Key Capabilities
- Enhanced Creative Writing: Specifically tuned for generating narratives and creative text.
- Improved Response Completion: Incorporates
LatitudeGames/Wayfarer-Large-70B-Llama-3.3to help the model produce more conclusive and less open-ended responses. - Llama-3.3 Foundation: Leverages the capabilities of the Llama-3.3 instruction-tuned models.
Merge Details
This iteration of Apocrypha integrates components from:
tdrussell/Llama-3-70B-Instruct-StorywriterLatitudeGames/Wayfarer-Large-70B-Llama-3.3nbeerbower/Llama3.1-Gutenberg-Doppel-70BTheDrummer/Fallen-Llama-3.3-70B-v1Doctor-Shotgun/L3.3-70B-Magnum-Diamond
The merge configuration used a select_topk value of 0.24 and specifies a Llama3 chat template. The model is intended for use cases requiring robust and coherent creative text generation.