ChamalyAI/gemma4-E2B-chamaliya
ChamalyAI/gemma4-E2B-chamaliya is a 5.1 billion parameter Gemma-4-E2B-it model, fine-tuned by ChamalyAI on approximately 60,000 samples of Northern Moroccan Darija (Jebli dialect). This model is specifically optimized for understanding and generating text in this unique dialect, preserving its distinct phonology, pronouns, Spanish loanwords, and diminutives. It is designed for applications requiring highly localized linguistic capabilities in Northern Moroccan Arabic.
Loading preview...
ChamalyAI/gemma4-E2B-chamaliya: Northern Moroccan Darija LLM
This model is a fully merged version of google/gemma-4-E2B-it, fine-tuned by ChamalyAI to specialize in the Northern Moroccan Darija (Jebli dialect). With 5.1 billion parameters and a 32768-token context length, it leverages LoRA fine-tuning on approximately 60,000 dialect-specific samples, with the adapter weights fused directly into the base model.
Key Capabilities & Features
- Dialectal Specialization: Proficient in Northern Moroccan Darija, including Jebli phonology (preserving Qaf), specific gender-neutral pronouns (نتينا), common Spanish loanwords (e.g., النيبيرا), and diminutives (e.g., عايل ستيتو).
- Merged Architecture: The LoRA adapter is fully merged into the base model, eliminating the need for separate adapter handling during deployment.
- RAFT Ready: Its merged nature makes it suitable for direct use as a policy model in Reward rAnked Fine-Tuning (RAFT workflows).
Ideal Use Cases
- Localized Content Generation: Creating text, dialogues, or narratives specifically for Northern Moroccan audiences.
- Dialectal Chatbots: Developing conversational AI agents that can interact naturally in the Jebli dialect.
- Linguistic Research: Studying the nuances and characteristics of Northern Moroccan Darija through a specialized language model.