Orca Mini v2 German 7b: German Language Adaptation

The jphme/orca_mini_v2_ger_7b is a specialized 7 billion parameter language model, fine-tuned by jphme from Pankaj Mathur's Orca Mini V2 7b. Its primary distinction is its optimization for the German language, achieved through additional training on a proprietary, handcrafted German dataset. The original Orca Mini V2 was built using 'explain-tuned' datasets, incorporating instruction and input from WizardLM, Alpaca, and Dolly-V2, and applying Orca Research Paper dataset construction methods.

Key Capabilities

• German Language Proficiency: Demonstrates enhanced ability to understand and generate German text, significantly outperforming the base model in German-specific interactions.
• Instruction Following: Inherits the strong instruction-following capabilities from its Orca Mini V2 base, which was trained on diverse instruction-tuned datasets.
• Experimental Development: Represents an ongoing effort to improve German language capabilities, with potential for future updates based on community interest.

Good for

• German Text Generation: Ideal for applications requiring natural and coherent text generation in German.
• German Language Understanding: Suitable for tasks involving comprehension and response to German instructions and queries.
• Research and Development: Useful for researchers and developers exploring multilingual LLM adaptations, particularly for low-resource or specialized language domains like German.

While currently limited by its experimental dataset and parameter count, the model provides a solid foundation for German-centric natural language processing tasks. A quantized GGML version for CPU inference is also available here.