| Property | Value |
|---|---|
| Base Model | Qwen2-7B |
| Format | GGUF |
| License | Apache License 2.0 |
| Author | Tifa-RP |
| HuggingFace | Link |
What is Tifa-7B-Qwen2-v0.1-GGUF?
Tifa is a high-performance language model distilled from a 220B parameter industrial model (VISIONSIC_220), specifically optimized for roleplay and dialogue generation. The model underwent extensive training with 400GB of novel data and 20GB of multi-turn dialogue data for fine-tuning.
Implementation Details
The model leverages the Qwen2-7B architecture and has been converted to GGUF format for compatibility with the Ollama framework. The training dataset comprises a carefully curated mix of Chinese roleplay (7.6%), English roleplay (4.2%), EFX industrial domain parameters (5.14%), and contemporary Q&A data generated from the original 220B model.
- Distillation from 220B to 7B parameters
- Specialized roleplay capabilities
- Industrial knowledge integration
- GGUF format optimization
Core Capabilities
- Multi-turn dialogue processing
- Character roleplay and scenario simulation
- EFX industrial knowledge integration
- High-quality literary creation
- Emotional expression and state tracking
Frequently Asked Questions
Q: What makes this model unique?
The model combines industrial knowledge with creative roleplay capabilities, distilled from a massive 220B parameter model into a more accessible 7B parameter version while maintaining sophisticated dialogue abilities.
Q: What are the recommended use cases?
The model excels in character roleplay scenarios, creative writing, multi-turn conversations, and industrial domain applications. It's particularly effective when used with f16 precision to preserve nuanced expressions.