Upload README.md with huggingface_hub

README.md

@@ -106,10 +106,33 @@ print(train_df.head())
 
 ## For Fine-tuning
 
-This dataset is designed for fine-tuning LLMs for spam detection. See the companion model:
+This dataset is designed for fine-tuning LLMs for spam detection. 
+
+### Training Details Used
+
+- **Epochs**: 5
+- **Batch Size**: 16
+- **Learning Rate**: 2e-5
+- **Optimizer**: AdamW
+- **Loss Function**: CrossEntropyLoss
+
+### Results
+
+| Metric | Score |
+|--------|-------|
+| Test Accuracy | ~95% |
+| Validation Accuracy | ~95%+ |
+
+### Companion Model
 
 **Model:** [mustafaege/spam-detector-gpt2-hf](https://huggingface.co/mustafaege/spam-detector-gpt2-hf)
 
+The model was trained following the book's Chapter 6.5 approach:
+- Base Model: openai-community/gpt2 (124M parameters)
+- Modified Layer: Added `out_head = nn.Linear(768, 2)` classification head
+- Unfrozen Layers: Last transformer block (h[-1]), Final LayerNorm (ln_f), Classification head (out_head)
+- Classification Method: Using the last token hidden state for prediction
+
 ## License
 
 The original UCI SMS Spam Collection dataset is available under the [UCI Machine Learning Repository](https://archive.ics.uci.edu/dataset/228/sms+spam+collection) terms.