dustyatx's picture
Update README.md
15e65a9 verified
# ModernBERT-BizClassifier - Business Relevance Classifier
## πŸ“Œ Overview
ModernBERT-BizClassifier is a **fine-tuned ModernBERT model** that classifies text from **any source** (social media, news, reports, etc.) and determines whether it is **useful for businesses**.
The model distinguishes between:
- **Business_Action_Direct**: Clearly relevant business information (e.g., financial reports, earnings, funding rounds).
- **Business_Intelligence_Indirect**: Text that implies business relevance (e.g., regulatory news, economic trends, emerging markets).
- **Not_Business_Relevant**: Text that contains no useful business insights.
## πŸ›  Model Details
| Hyperparameter | Value |
|-----------------------|----------------|
| **Base Model** | `ModernBERT-base` |
| **Max Context Length** | `8192 tokens` |
| **Batch Size** | `32` |
| **Learning Rate** | `1e-5` |
| **Weight Decay** | `0.005` |
| **Epochs** | `2.0` |
| **Warmup Ratio** | `0.1` |
| **Precision Used** | `BF16` if CUDA available |
| **Problem Type** | `Multi-label classification` |
## πŸ“Š Evaluation Results
| Metric | Score |
|-----------------------------|--------|
| **Final Validation Loss** | `0.2940` |
| **Micro F1 Score** | `0.8049` |
| **Macro F1 Score** | `0.8042` |
| **Evaluation Runtime** | `3.48 seconds` |
| **Samples Processed Per Second** | `239.09` |
| **Steps Processed Per Second** | `7.48` |
## πŸ— Model Card
- **Tags**: `business-classification`, `BERT`, `financial-analysis`, `text-classification`
```python
{
1: "Business_Action_Direct",
2: "Business_Intelligence_Indirect",
3: "Not_Business_Relevant"
}
```