Update README.md

README.md

@@ -8,15 +8,104 @@ tags:
 - trl
 license: apache-2.0
 language:
-- en
+- ar
 ---
 
-# Uploaded  model
+# Qwen2 VL - Arabic OCR Fine-Tuned Model
+
+## Model Overview
+This model is a fine-tuned version of [Qwen2 VL](https://huggingface.co/Qwen/Qwen2-VL) on an Arabic OCR dataset. It is optimized to perform Arabic Optical Character Recognition (OCR) for full-page text.
+
+## Model Details
+- **Base Model**: Qwen2 VL
+- **Fine-tuning Dataset**: Arabic OCR dataset
+- **Objective**: Extract full-page Arabic text with high accuracy
+- **Languages**: Arabic
+- **Tasks**: OCR (Optical Character Recognition)
+
+## Evaluation Results
+
+The fine-tuned model outperforms the base model significantly in terms of Character Error Rate (CER), Word Error Rate (WER), and BLEU score.
+
+### Fine-Tuned Model Performance
+- **Word Error Rate (WER)**: `0.0675`
+- **Character Error Rate (CER)**: `0.0193`
+- **BLEU Score**:
+  - BLEU: `0.8596`
+  - Precision @1: `93.95%`
+  - Precision @2: `88.55%`
+  - Precision @3: `83.82%`
+  - Precision @4: `79.52%`
+
+### Base Model Performance
+- **Word Error Rate (WER)**: `1.3435`
+- **Character Error Rate (CER)**: `1.1915`
+- **BLEU Score**:
+  - BLEU: `0.2007`
+  - Precision @1: `26.85%`
+  - Precision @2: `21.65%`
+  - Precision @3: `18.13%`
+  - Precision @4: `15.39%`
+
+## Performance Comparison Charts
+
+### WER & CER Comparison
+```python
+import matplotlib.pyplot as plt
+
+categories = ["WER", "CER"]
+base_values = [1.3435, 1.1915]
+fine_tuned_values = [0.0675, 0.0193]
+
+x = range(len(categories))
+plt.bar(x, base_values, width=0.4, label="Base Model", color='r', align='center')
+plt.bar(x, fine_tuned_values, width=0.4, label="Fine-Tuned Model", color='g', align='edge')
+
+plt.xticks(x, categories)
+plt.ylabel("Error Rate")
+plt.title("WER & CER Comparison")
+plt.legend()
+plt.show()
+```
+
+### BLEU Score Comparison
+```python
+categories = ["BLEU", "Precision @1", "Precision @2", "Precision @3", "Precision @4"]
+base_bleu = [0.2007, 26.85, 21.65, 18.13, 15.39]
+fine_tuned_bleu = [0.8596, 93.95, 88.55, 83.82, 79.52]
+
+x = range(len(categories))
+plt.bar(x, base_bleu, width=0.4, label="Base Model", color='r', align='center')
+plt.bar(x, fine_tuned_bleu, width=0.4, label="Fine-Tuned Model", color='g', align='edge')
+
+plt.xticks(x, categories)
+plt.ylabel("Score (%)")
+plt.title("BLEU Score & Precision Comparison")
+plt.legend()
+plt.show()
+```
+
+## How to Use
+You can load this model using the `transformers` library:
+
+```python
+from transformers import AutoModel, AutoProcessor
+import torch
+
+model_name = "your-model-name"
+model = AutoModel.from_pretrained(model_name)
+processor = AutoProcessor.from_pretrained(model_name)
+
+image = "path/to/your/image.jpg"
+inputs = processor(images=image, return_tensors="pt")
+outputs = model(**inputs)
+```
+
+## License
+This model follows the licensing terms of the original Qwen2 VL model. Please review the terms before using it commercially.
+
+## Citation
+If you use this model in your research or application, please cite it appropriately.
 
-- **Developed by:** oddadmix
-- **License:** apache-2.0
-- **Finetuned from model :** unsloth/qwen2-vl-2b-instruct-unsloth-bnb-4bit
 
-This qwen2_vl model was trained 2x faster with [Unsloth](https://github.com/unslothai/unsloth) and Huggingface's TRL library.
 
-[<img src="https://raw.githubusercontent.com/unslothai/unsloth/main/images/unsloth%20made%20with%20love.png" width="200"/>](https://github.com/unslothai/unsloth)