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.