Mythalion-13B-GPTQ
| Property | Value |
|---|---|
| Parameter Count | 2.03B (Quantized) |
| Model Type | LLaMA2-based |
| License | LLaMA2 |
| Quantization | GPTQ (Multiple Options) |
What is Mythalion-13B-GPTQ?
Mythalion-13B-GPTQ is a quantized version of the collaborative effort between PygmalionAI and Gryphe, combining Pygmalion-2 13B and MythoMax L2 13B models. This GPTQ-quantized variant offers various compression options while maintaining performance, making it more accessible for users with different hardware capabilities.
Implementation Details
The model comes in multiple GPTQ configurations, ranging from 4-bit to 8-bit quantization with different group sizes (32g, 64g, 128g) and Act Order options. The implementation supports both Alpaca and Pygmalion formatting styles, making it versatile for different use cases.
- Multiple quantization options (4-bit and 8-bit variants)
- Various group sizes for VRAM optimization
- Compatible with ExLlama (4-bit versions)
- Supports both Transformers and AutoGPTQ frameworks
Core Capabilities
- Optimized for roleplay and chat interactions
- Supports multi-turn conversations
- Enhanced fictional writing capabilities
- Flexible prompt formatting support
- Multiple dataset training (including PIPPA, OpenOrca, Claude multiround chat)
Frequently Asked Questions
Q: What makes this model unique?
This model combines the strengths of Pygmalion-2 and MythoMax, offering superior performance in roleplay and chat scenarios while providing multiple quantization options for different hardware requirements.
Q: What are the recommended use cases?
The model is primarily designed for fictional writing and entertainment purposes, excelling in roleplay and chat applications. It's particularly well-suited for use with platforms like SillyTavern with specific recommended generation settings.
