# SIB200 CDA Model with Gemma

This model was trained on the SIB200 dataset using Counterfactual Data Augmentation (CDA) with counterfactuals generated by Gemma.

## Training Parameters

- **Dataset**: SIB200
- **Mode**: CDA
- **Selection Model**: Gemma
- **Selection Method**: Random
- **Train Size**: 700 examples
- **Epochs**: 20
- **Batch Size**: 8
- **Effective Batch Size**: 32 (batch_size * gradient_accumulation_steps)
- **Learning Rate**: 8e-06
- **Patience**: 8
- **Max Length**: 192
- **Gradient Accumulation Steps**: 4
- **Warmup Ratio**: 0.1
- **Weight Decay**: 0.01
- **Optimizer**: AdamW
- **Scheduler**: cosine_with_warmup
- **Random Seed**: 42

## Performance

- **Overall Accuracy**: 77.95%
- **Overall Loss**: 0.0237

### Language-Specific Performance
- **English (EN)**: 86.87%
- **German (DE)**: 86.87%
- **Arabic (AR)**: 49.49%
- **Spanish (ES)**: 89.90%
- **Hindi (HI)**: 83.84%
- **Swahili (SW)**: 70.71%

## Model Information

- **Base Model**: bert-base-multilingual-cased
- **Task**: Topic Classification
- **Languages**: 6 languages (EN, DE, AR, ES, HI, SW)