File size: 905 Bytes
f1bf05e |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 |
# MariaOls/RussianDMRecognizer_dual
Dual-headed model (BIO token tagging + sentence-level DM detection) fine-tuned from **viktoroo/sberbank-rubert-base-collection3**.
Ready for inference with `AutoModel(..., trust_remote_code=True)` or your `main.py` (sentence by sentence).
## Eval metrics
| metric | value |
|---|---|
| eval_bio_accuracy | 0.993706 |
| eval_bio_precision | 0.925270 |
| eval_bio_recall | 0.936318 |
| eval_bio_f1 | 0.930762 |
| eval_cls_accuracy | 0.951408 |
| eval_cls_precision | 0.966135 |
| eval_cls_recall | 0.965174 |
| eval_cls_f1 | 0.965655 |
## Usage (Python)
```python
from transformers import AutoModel, AutoTokenizer
m = AutoModel.from_pretrained("MariaOls/RussianDMRecognizer_dual", trust_remote_code=True)
t = AutoTokenizer.from_pretrained("MariaOls/RussianDMRecognizer_dual", use_fast=True)
# m(...) devuelve dict con "logits": (logits_tok [B,L,3], logits_seq [B,2])
|