ahmadabdulnasir/NaijaPidgin-Qwen3-4B
The ahmadabdulnasir/NaijaPidgin-Qwen3-4B is a 4 billion parameter Qwen3-based causal language model fine-tuned by Ahmad Abdulnasir. It specializes in understanding and generating Nigerian Pidgin English, offering capabilities like conversational Pidgin, Pidgin-English translation, and knowledge of Nigerian culture. This model is optimized for applications requiring fluent interaction in Nigerian Pidgin English, including code-switching with standard English.
Loading preview...
NaijaPidgin-Qwen3-4B: Nigerian Pidgin English LLM
This model is a fine-tuned version of the Qwen3-4B base model, specifically adapted by Ahmad Abdulnasir to understand and generate Nigerian Pidgin English (Naija). It was trained using QLoRA (4-bit) with Unsloth on the ahmadabdulnasir/naijaPidgin dataset, leveraging a Google Colab G4 GPU.
Key Capabilities
- Conversational Pidgin English: Engages in natural dialogue using Nigerian Pidgin.
- Pidgin ↔ English Translation: Facilitates translation between Pidgin and standard English.
- Cultural Understanding: Incorporates knowledge of Nigerian culture, proverbs, and everyday advice.
- Code-switching: Seamlessly switches between Pidgin and English within conversations.
- Thinking Mode: Utilizes Qwen3's inherent reasoning capabilities for complex questions when enabled.
Good For
- Developing chatbots or virtual assistants for Nigerian audiences.
- Translating content to or from Nigerian Pidgin English.
- Applications requiring culturally relevant responses in a Nigerian context.
- Exploring the nuances of code-switching in a low-resource language setting.