Upload README.md

README.md

@@ -12,4 +12,33 @@ metrics:
   - qwk
   - rmse
 pipeline_tag: text-classification
----
+---
+
+# IELTS Writing Task 2 Essay Scorer (DeBERTa-v3-base)
+
+This repository contains a custom **Prompt-Aware IELTS Essay Scorer** fine-tuned on DeBERTa-v3-base. It simultaneously predicts the 4 analytical criteria of IELTS Writing Task 2 criteria and computes the overall band score.
+
+## 🔗 Quick Links & Resources
+* **Dataset:** [View Clean Dataset](https://huggingface.co/star092304/ielts-writing-task2-debertav3base/tree/main/data)
+* **Training Logs:** [View training_history.csv](https://huggingface.co/star092304/ielts-writing-task2-debertav3base/blob/main/training_history.csv)
+* **Training Notebook:** [src/deberta-v3-base-ielts-aes.ipynb](https://huggingface.co/star092304/ielts-writing-task2-debertav3base/blob/main/src/deberta-v3-base-ielts-aes.ipynb)
+* **Inference Notebook:** [src/IELTS_DeBERTa_Inference.ipynb](https://huggingface.co/star092304/ielts-writing-task2-debertav3base/blob/main/src/IELTS_DeBERTa_Inference.ipynb)
+
+---
+
+## Model Capabilities
+The model takes both the **Writing Prompt** and the **Student's Essay** as inputs, focuses specifically on the essay body using an attention pooling layer, and grades the text across 4 core traits (scaled 0-9):
+* **TA** (Task Achievement)
+* **CC** (Coherence and Cohesion)
+* **LR** (Lexical Resource)
+* **GRA** (Grammatical Range and Accuracy)
+* **OverallBand** (Calculated average rounded to the nearest 0.5)
+
+### Training Curves
+![Training curves](training_curves.png)
+
+---
+
+## License
+
+This project is licensed under the MIT License.