StableLM 2 12B Chat
| Property | Value |
|---|---|
| Parameter Count | 12.1B |
| License | StabilityAI Non-Commercial Research Community License |
| Paper | Stable LM 2 Chat Technical Report |
| Language | English |
| Training Datasets | 10 datasets including UltraChat, MetaMathQA, WizardLM |
What is stablelm-2-12b-chat?
StableLM 2 12B Chat is an advanced language model developed by Stability AI, utilizing Direct Preference Optimization (DPO) for instruction tuning. The model represents a significant advancement in conversational AI, achieving impressive benchmark scores and offering versatile functionality including chat and function calling capabilities.
Implementation Details
The model is built on the transformer decoder architecture and supports the ChatML format. It requires transformers>=4.40.0 and can be easily implemented using the Hugging Face transformers library. The model supports both regular chat interactions and advanced function calling features, making it suitable for various applications.
- Supports BF16 tensor type for efficient processing
- Implements ChatML format for structured conversations
- Includes function calling capabilities for extended functionality
- Achieves 8.15 score on MT-Bench, competing with larger models
Core Capabilities
- High-quality text generation and conversation
- Function calling for integrated system operations
- Strong performance on multiple benchmarks including ARC Challenge and HellaSwag
- Efficient processing with optimized architecture
Frequently Asked Questions
Q: What makes this model unique?
The model stands out for its impressive performance despite its relatively moderate size, achieving scores comparable to larger models like Mixtral-8x7B-Instruct. It also features comprehensive function calling capabilities and is trained on a diverse set of high-quality datasets.
Q: What are the recommended use cases?
The model is ideal for chat-like applications, research purposes, and scenarios requiring function calling capabilities. However, it's recommended to implement proper input/output safeguards and evaluate the model's safety performance for specific use cases.
