Fine-tuning LLaMA on Classical Arabic Poetry
Fine-tuning and training LLaMA model to understand and generate classical metered and rhymed Arabic poetry.
75% Filled
Cohort Details
Arabic Poetry Fine-tuning & Training Project
This project aims to train the LLaMA 3 model on the vast heritage of classical Arabic poetry, enabling it to understand poetic meters (Al-Buhur), weights, and rhymes, and generate high-quality metered verses while preserving eloquence and cultural aesthetics.
Developer Activities & Implementation Status
Phase 1: Dataset Collection & Curation
CompletedAssembling Arabic classical poetry books, cleaning records, and formatting as instruction-tuning QA pairs.
Metered Verses Cured50,000 / 50,000
Digital / Remote50,000 developer
Phase 2: Fine-Tuning & Quantization
OngoingUsing LoRA techniques to train the model on cloud GPU instances and evaluating outputs.
GPU Training Hours80 / 100
Digital / Remote80 developer
Cohort Gallery & Demo Documentation
Join this Cohort
Registeredof capacity
75100
75% Filled
Your participation directly contributes to open-source development and enriches the technical content of our communities.
Join Cohort