Add pipeline_tag and proper model card for Inference API

README.md

@@ -1,22 +1,75 @@
-# Quran Q&A Model (Interrupted Training)
+---
+license: apache-2.0
+pipeline_tag: text-generation
+tags:
+- transformers
+- text-generation
+- arabic
+- quran
+- islamic
+- safetensors
+language:
+- ar
+library_name: transformers
+---
 
-## Status
-Training was interrupted but the model achieved excellent performance:
-- **ROUGE-1: 0.9989 (99.89%)**
-- Model is fully functional for Q&A tasks
+# QuranPlus
 
-## Model Details
-- Base: google/flan-t5-small
-- Date: 2025-07-11
-- Training interrupted at epoch Unknown
+QuranPlus is a language model trained for Islamic and Quranic text generation.
+
+## Model Description
+
+This model is designed to generate text related to Islamic teachings and Quranic content in Arabic.
 
 ## Usage
-Run test_model.py to test the model:
-```bash
-python test_model.py
+
+```python
+from transformers import AutoTokenizer, AutoModelForCausalLM
+
+# Load model and tokenizer
+tokenizer = AutoTokenizer.from_pretrained("justdeen/QuranPlus")
+model = AutoModelForCausalLM.from_pretrained("justdeen/QuranPlus")
+
+# Generate text
+input_text = "ما هو الإسلام؟"
+inputs = tokenizer(input_text, return_tensors="pt")
+
+with torch.no_grad():
+    outputs = model.generate(**inputs, max_length=100, do_sample=True)
+
+generated_text = tokenizer.decode(outputs[0], skip_special_tokens=True)
+print(generated_text)
 ```
 
-## Files
-- final_model/: The trained model and tokenizer
-- training_summary.json: Training metrics and configuration
-- test_model.py: Example usage script with test cases
+## Inference API
+
+You can also use this model via the Hugging Face Inference API:
+
+```python
+import requests
+
+API_URL = "https://api-inference.huggingface.co/models/justdeen/QuranPlus"
+headers = {"Authorization": f"Bearer {YOUR_HF_TOKEN}"}
+
+def query(payload):
+    response = requests.post(API_URL, headers=headers, json=payload)
+    return response.json()
+
+output = query({
+    "inputs": "ما هو الإسلام؟",
+})
+```
+
+## Training Details
+
+This model was trained on Islamic and Quranic texts to provide accurate and contextually appropriate responses about Islamic teachings.
+
+## Limitations
+
+- The model is specifically trained for Islamic content
+- Responses should be verified by Islamic scholars for religious accuracy
+- May not perform well on non-Islamic topics
+
+## License
+
+Apache 2.0