---
language:
- en
- es
- pt
- ar
- bn
- hi
license: mit
tags:
- football
- soccer
- multilingual
- gpt
- pytorch
- safetensors
pipeline_tag: text-generation
---

# ⚽ SOCCER BERT

A multilingual football-specific BERT model trained from scratch.

## Model Details

| Property | Value |
|----------|-------|
| Parameters | 159M |
| Architecture | Decoder-only Transformer |
| Layers | 10 |
| Attention Heads | 8 |
| Embedding Dim | 512 |
| Context Length | 512 tokens |
| Vocab Size | 250,002 (XLM-RoBERTa) |
| Training Phase | 2 of 3 |
| Dataset | 3.72 GB football + wikipedia |
| Format | Safetensors |
| Status | Private (Phase 3 SFT pending) |

## Languages
English, Spanish, Portuguese, Arabic, Bengali, Hindi


## Citation
```bibtex
@misc{Soccer-BERT-2026,
  author = {Tareq},
  title  = {Football GPT V2},
  year   = {2025},
  url    = {https://huggingface.co/tareq052/football-gpt-v2}
}
```