# Model Card: SDGs Classification

## Model Overview

**Model Name:** sdd-sdgs  
**Base Model:** `indobenchmark/indobert-base-p2`  
**Task:** Multi-class SDG classification (18-class)  
**Language:** Indonesian

---

## Model Description

Fine-tuned IndoBERT for classifying Indonesian text against UN Sustainable Development Goals.

**Classes:**
SDG0 (Non-SDG), SDG1-SDG17 (UN Sustainable Development Goals)

---

## Performance Metrics

| Metric | Test Public | Test Internal |
|---|---|---|
| Accuracy | 0.7919 | 1.0000 |
| Macro F1 | 0.7445 | 1.0000 |
| Latency (mean) | 8.71 ms | - |
| Model Size | 474.8 MB | - |

---

## Usage

### Load Model

```python
from transformers import AutoTokenizer, AutoModelForSequenceClassification
import torch

model_name = "AzrilFahmiardi/sdd-sdgs"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForSequenceClassification.from_pretrained(model_name)

device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
model = model.to(device)

SDG_DICT = {
    "SDG0": "Non-SDG", "SDG1": "No Poverty", "SDG2": "Zero Hunger",
    "SDG3": "Good Health", "SDG4": "Quality Education", "SDG5": "Gender Equality",
    "SDG6": "Clean Water", "SDG7": "Clean Energy", "SDG8": "Decent Work",
    "SDG9": "Industry Innovation", "SDG10": "Reduced Inequalities",
    "SDG11": "Sustainable Cities", "SDG12": "Responsible Consumption",
    "SDG13": "Climate Action", "SDG14": "Life Below Water", "SDG15": "Life on Land",
    "SDG16": "Peace Justice", "SDG17": "Partnerships"
}
```

### Inference

```python
def classify_sdg(text: str) -> dict:
    inputs = tokenizer(text, return_tensors="pt", truncation=True, max_length=256).to(device)
    
    with torch.no_grad():
        outputs = model(**inputs)
        logits = outputs.logits
    
    probabilities = torch.softmax(logits, dim=-1)[0].cpu()
    predicted_class = logits.argmax(-1).item()
    predicted_label = model.config.id2label[predicted_class]
    confidence = probabilities[predicted_class].item()
    
    return {
        "sdg_code": predicted_label,
        "sdg_name": SDG_DICT[predicted_label],
        "confidence": confidence
    }

# Example
text = "Pemerintah meluncurkan program pengentasan kemiskinan di daerah terpencil."
result = classify_sdg(text)
print(f"SDG: {result['sdg_code']} - {result['sdg_name']} ({result['confidence']:.2%})")
```

### Output Format

```json
{
  "sdg_code": "SDG1",
  "sdg_name": "No Poverty",
  "confidence": 0.8756
}
```

---

## Input/Output

| Parameter | Type | Example |
|---|---|---|
| **Input** | str | Indonesian text (policy, news), max 256 tokens |
| **Output** | dict | `{"sdg_code": "SDG1", "sdg_name": "No Poverty", "confidence": 0.88}` |