Dusk_Rainbow
| Property | Value |
|---|---|
| Parameter Count | 8.03B |
| License | Apache 2.0 |
| Format | Safetensors FP16 |
| Language | English |
What is Dusk_Rainbow?
Dusk_Rainbow is a specialized language model designed for creative writing and general tasks with minimal censorship. Built on the EtherealRainbow v0.3 8B base model, it has been fine-tuned on 16M tokens of carefully curated data to provide exceptional creative writing capabilities while maintaining low censorship levels (rated 9.1/10 on the uncensored scale).
Implementation Details
The model utilizes a unique training approach that avoids using ChatGPT or Claude-generated data to maintain authenticity and reduce repetitive patterns. It's available in multiple quantizations including FP16, GGUF, EXL2, and specialized FP8 versions for different deployment scenarios.
- Built on EtherealRainbow base model architecture
- Fine-tuned on 16M tokens of specialized data
- Available in multiple quantization formats
- Supports both ChatML and Llama-3 instruction templates
Core Capabilities
- Exceptional creative writing with strict paragraph adherence
- Strong instruction following abilities
- Minimal censorship for unrestricted creative expression
- Versatile deployment options across different platforms
- Benchmark performance: 18.49 average across standard metrics
Frequently Asked Questions
Q: What makes this model unique?
The model's unique training approach avoiding API-generated content, combined with its specialized focus on creative writing and low censorship, makes it particularly effective for storytelling and creative tasks while maintaining authentic outputs.
Q: What are the recommended use cases?
Dusk_Rainbow excels in creative writing tasks, story generation, and general content creation where unrestricted expression is desired. It's particularly effective when used with Debug Deterministic or min_p settings for consistent results.
