KaraKaraWarehouse/LLENN-v0.75-Qwen2.5-72b
KaraKaraWitch/LLENN-v0.75-Qwen2.5-72b is a 72.7 billion parameter merged language model based on the Qwen2.5 architecture, developed by KaraKaraWitch. This model integrates components from several Qwen-based 72B models, including Rombos-LLM-V2.5, Dracarys2, EVA-Qwen2.5, Chronos-Platinum, and banana-2-b. It is designed for general text generation tasks, supporting a wide array of languages including Chinese, English, French, Spanish, and more, with a notable context length of 131072 tokens.
Loading preview...
LLENN-v0.75-Qwen2.5-72b Overview
LLENN-v0.75-Qwen2.5-72b is a 72.7 billion parameter merged language model created by KaraKaraWitch, building upon the Qwen2.5 architecture. This iteration, v0.75, refines a previous version by adjusting its "claude vibes" through the removal of a component, while retaining the core merge structure due to the absence of new models for integration.
Key Capabilities & Features
- Merged Architecture: Combines the strengths of five distinct Qwen-based 72B models:
- Multilingual Support: Capable of processing and generating text in numerous languages, including Chinese, English, French, Spanish, Portuguese, German, Italian, Russian, Japanese, Korean, Vietnamese, Thai, and Arabic.
- High Context Length: Features a substantial context window of 131072 tokens, allowing for extensive conversational and document processing.
- Merge Method: Utilizes the
model_stockmerge method with normalization, based on theQwen/Qwen2.5-72Bas the base model, and processed inbfloat16dtype.
Recommended Usage
- Prompt Format: Primarily designed to work with ChatML format.
- Sampler Settings: Users can experiment with suggested sampler settings for optimal output, such as Temperature 1.2, Min P 0.07, and Repetition Penalty 1.1, or an alternative of Temperature 1.1, Top P 0.98, and Min P 0.05.
Top 3 parameter combinations used by Featherless users for this model. Click a tab to see each config.