NSFW_13B_sft
| Property | Value |
|---|---|
| Base Model | Baichuan-13B-Base |
| Language | Chinese |
| License | Apache 2.0 |
| Training Framework | LLaMA-Efficient-Tuning |
What is NSFW_13B_sft?
NSFW_13B_sft is an instruction-tuned language model based on Baichuan-13B-Base, specifically designed for Chinese text generation. The model has been fine-tuned using the instruct_nsfw_cn dataset and demonstrates improved performance across various CMMLU benchmarks compared to its base model.
Implementation Details
The model leverages the LLaMA-Efficient-Tuning framework for training and can be easily implemented using the Transformers library. It supports text generation with streaming capabilities and can be deployed either through Python code or via a CLI interface.
- Built on PyTorch framework
- Implements efficient instruction tuning techniques
- Supports text streaming for real-time generation
- Includes custom tokenizer implementation
Core Capabilities
- Outperforms base model in CMMLU benchmarks (Overall score: 50.42 vs 48.86)
- Excels in China-specific knowledge (51.84%) and Social Sciences (55.55%)
- Supports long-form text generation with up to 1024 tokens
- Specialized in Chinese language processing
Frequently Asked Questions
Q: What makes this model unique?
The model stands out for its specialized instruction tuning on Chinese NSFW content while maintaining strong performance across general knowledge domains. It notably outperforms the original Baichuan-13B-Chat model in several CMMLU categories.
Q: What are the recommended use cases?
The model is primarily designed for Chinese text generation tasks and performs particularly well in humanities, social sciences, and China-specific knowledge domains. It's important to note that the model contains NSFW capabilities and should be used responsibly.
