Hugging Face Overview
KaraKaraWitch/Llama-ProgressPushDoll-3.3-70Bees is a 70 billion parameter language model built upon the Llama-3.3-70B-Instruct base. Developed by KaraKaraWitch, this model was created using the Model Stock merge method to combine the strengths of multiple Llama-3.3-70B variants. The goal of this merge was to produce a model that performs well "out of the box" without extensive parameter wrangling.
Merge Details
This model integrates nine different Llama-3.3-70B based models, including:
- KaraKaraWitch/Llama-MiraiFanfare-2-3.3-70B
- Undi95/Sushi-v1.4
- Nohobby/L3.3-Prikol-70B-v0.2
- Sao10K/L3.3-70B-Euryale-v2.3
- TheDrummer/Anubis-70B-v1
- EVA-UNIT-01/EVA-LLaMA-3.33-70B-v0.1
- nitky/Llama-3.3-SuperSwallowX-70B-Instruct-v0.1
- Blackroot/Mirai-3.0-70B
- Sao10K/70B-L3.3-Cirrus-x1
The merge utilized a normalize: true parameter configuration and bfloat16 dtype, aiming for a cohesive and stable performance profile.
Prompt Formats
The model is compatible with both ChatML and L3 chat prompt formats, offering flexibility for integration into various conversational AI systems.
Intended Use
Llama-ProgressPushDoll-3.3-70Bees is designed for users seeking a robust 70B Llama-3.3-based model that delivers solid performance across general conversational tasks, minimizing the need for extensive fine-tuning or complex parameter adjustments.