Update README.md

README.md

@@ -1,3 +1,64 @@
----
-license: mit
----
+---
+license: mit
+language:
+- en
+tags:
+- text2text-generation
+- sql
+- text-to-sql
+- gemma
+- fine-tuned
+- database
+- nlp
+base_model: google/gemma-7b
+datasets:
+- estu-research/sql-training-dataset
+metrics:
+- accuracy
+- exact_match
+library_name: transformers
+pipeline_tag: text2text-generation
+---
+
+# Gemma-7B SQL Expert (Fine-Tuned)
+
+Fine-tuned version of Google's Gemma-7B model for converting natural language questions to SQL queries.
+
+## Model Details
+
+- **Base Model**: [google/gemma-7b](https://huggingface.co/google/gemma-7b)
+- **Fine-tuned by**: ESTU Research Team (Kulalı, Aydın, Alhan, Fidan)
+- **Institution**: Eskisehir Technical University
+- **Project**: TÜBİTAK 2209-A Research
+- **License**: MIT
+- **Language**: English
+- **Task**: Natural Language to SQL Translation
+
+## Performance
+
+- **Execution Accuracy**: 76.0%
+- **Exact Match**: 65.4%
+- **Average Latency**: 500ms
+- **Model Size**: 14.1 GB (full) / 183 MB (LoRA adapters)
+
+## Training Details
+
+### Training Data
+- **Dataset**: [estu-research/sql-training-dataset](https://huggingface.co/datasets/estu-research/sql-training-dataset)
+- **Examples**: 1,000+ natural language to SQL pairs
+- **Domain**: Sales database queries (customers, orders, products, employees)
+
+### Training Configuration
+```python
+{
+  "base_model": "google/gemma-7b",
+  "method": "LoRA",
+  "rank": 16,
+  "alpha": 32,
+  "dropout": 0.05,
+  "target_modules": ["q_proj", "k_proj", "v_proj", "o_proj"],
+  "epochs": 3,
+  "batch_size": 8,
+  "learning_rate": 1.5e-4,
+  "training_time": "10.8 hours (A100 GPU)"
+}