This model is a domain-specific fine-tuned version of Qwen2.5-7B-Instruct, optimized for Islamic Jurisprudence (Fiqh) and Aqidah based on the Shamela Library corpus.
📊 Training Statistics & Metrics
Metric
Initial Value
Final Value
Improvement
Cross-Entropy Loss
3.8489
2.0326
47.19%
Perplexity (PPL)
46.94
7.63
83.74%
📚 Dataset Details
Source: Shamela Full Merged (Subset of 230 Scientific Titles).
Total Processed Tokens:24,576,000 Tokens.
Reference List: See Azhar_Model_v0.3_Dataset_Metadata.xlsx in the files section.
📉 Visual Analysis
Training Metrics
🎓 Citation & Thesis
Developed by MSC. Shamil Al-Mohammedi as part of a doctoral research project.
