| Property | Value |
|---|---|
| Parameter Count | 7.24B |
| Base Model | LeoLM/leo-mistral-hessianai-7b |
| License | Apache 2.0 |
| Supported Languages | German, English |
| Tensor Type | BF16 |
What is DiscoLM_German_7b_v1?
DiscoLM German 7b is a specialized language model developed by DiscoResearch, designed primarily for German language applications while maintaining strong English language capabilities. It's built on the Mistral architecture and has undergone both SFT (Supervised Fine-Tuning) and DPO (Direct Preference Optimization) training phases to enhance its performance.
Implementation Details
The model implements the ChatML format for prompting, making it compatible with OpenAI endpoints and various inference libraries. It features special capabilities for retrieval applications and experimental function calling support.
- Utilizes ChatML prompt format for structured conversations
- Supports system prompts for enhanced steerability
- Includes specialized retrieval format for RAG applications
- Features experimental function calling capabilities
Core Capabilities
- Advanced German language understanding and generation
- Strong performance in German-English translation tasks
- Competitive performance against GPT-3.5-turbo in reasoning tasks
- RAG-optimized retrieval format support
- Multi-turn conversation handling
Frequently Asked Questions
Q: What makes this model unique?
The model's primary distinction is its optimization for German language tasks, with reported quality sometimes exceeding GPT-4 for German content. It's designed as a reliable alternative to proprietary models for everyday German language applications.
Q: What are the recommended use cases?
The model excels in German text generation, translation tasks, and general conversation. It's particularly suitable for RAG applications and business use cases requiring German language proficiency. However, it may not match larger models in complex reasoning, mathematics, or coding tasks.