# XLM-RoBERTa Multi-Head (TR) — Intent + NER

- Base: `xlm-roberta-base`
- Tasks: Intent classification (60 sınıf), NER (BIO)
- Dosyalar: `pytorch_model.bin`, `config.json`, `tokenizer.*`, `label_schemes.json`, `modeling_xlmr_multihead.py`

## Kullanım

```python
from transformers import AutoTokenizer
from modeling_xlmr_multihead import XLMRMultiHead
import torch, json

ckpt = "celalkartoglu/xlmr-multihead-tr"

tok = AutoTokenizer.from_pretrained(ckpt)
model = XLMRMultiHead(n_intent=60, n_ner=111) 
model.load_state_dict(torch.load("pytorch_model.bin", map_location="cpu"))
model.eval()