akarki15/nepali-rapper-merged
The akarki15/nepali-rapper-merged is an 8 billion parameter Qwen3-based causal language model, fine-tuned by akarki15 to generate Nepali rap verses and conversations. This model integrates a LoRA adapter, specifically trained on approximately 50 Nepali rapper conversations, directly into the base weights, making it ready for immediate use without PEFT. It excels at producing content in mixed Nepali (Devanagari and Romanized) and English, incorporating Nepali slang and hip-hop lingo, making it ideal for creative text generation in a specific cultural and linguistic style.
Loading preview...
Overview
akarki15/nepali-rapper-merged is an 8 billion parameter language model built upon the Qwen3-8B architecture. It has been specifically fine-tuned by akarki15 to generate text in the style of a Nepali rapper, incorporating Nepali slang, hip-hop lingo, and a mix of Devanagari and Romanized Nepali with English.
Key Capabilities
- Nepali Rap Generation: Excels at creating rap verses, freestyles, and diss tracks with authentic Nepali street and hip-hop cultural nuances.
- Multilingual Output: Seamlessly blends Nepali (Devanagari and Romanized) and English in its responses.
- Character Emulation: Designed to converse with the "swag and attitude" of a Nepali rapper, as defined by its system prompt.
- Integrated Adapter: The
nepali-rapper-loraadapter is merged into the base model, allowing for direct use without requiring PEFT (Parameter-Efficient Fine-Tuning).
Training Details
The model was fine-tuned using LoRA (r=16, alpha=16) on approximately 50 multi-turn conversations in ShareGPT format. These conversations covered topics such as verse/freestyle generation, battle rap, Nepal-themed raps, and casual rapper chat. Training was conducted for 3 epochs using Unsloth and TRL SFTTrainer, taking about 10-15 minutes on a Google Colab T4 GPU.
Good For
- Creative writing applications requiring Nepali rap content.
- Chatbots or interactive experiences simulating a Nepali rapper persona.
- Generating culturally specific text with mixed language elements.