File size: 1,780 Bytes
3ed31d6 62fe5ad |
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 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 |
---
language:
- ru
tags:
- token-classification
license: apache-2.0
widget:
- text: Ёпта, меня зовут придурок и я живу в жопе
---
# RuBERTConv Toxic Editor
## Model description
Tagging model for detoxification based on [rubert-base-cased-conversational](https://huggingface.co/DeepPavlov/rubert-base-cased-conversational).
4 possible classes:
- Equal = save tokens
- Replace = replace tokens with mask
- Delete = remove tokens
- Insert = insert mask before tokens
Use in pair with [mask filler](https://huggingface.co/IlyaGusev/sber_rut5_filler).
## Intended uses & limitations
#### How to use
Colab: [link](https://colab.research.google.com/drive/1NUSO1QGlDgD-IWXa2SpeND089eVxrCJW)
```python
import torch
from transformers import AutoTokenizer, pipeline
tagger_model_name = "IlyaGusev/rubertconv_toxic_editor"
device = "cuda" if torch.cuda.is_available() else "cpu"
device_num = 0 if device == "cuda" else -1
tagger_pipe = pipeline(
"token-classification",
model=tagger_model_name,
tokenizer=tagger_model_name,
framework="pt",
device=device_num,
aggregation_strategy="max"
)
text = "..."
tagger_predictions = tagger_pipe([text], batch_size=1)
sample_predictions = tagger_predictions[0]
print(sample_predictions)
```
## Training data
- Dataset: [russe_detox_2022](https://github.com/skoltech-nlp/russe_detox_2022/tree/main/data)
## Training procedure
- Parallel corpus convertion: [compute_tags.py](https://github.com/IlyaGusev/rudetox/blob/main/rudetox/marker/compute_tags.py)
- Training script: [train.py](https://github.com/IlyaGusev/rudetox/blob/main/rudetox/marker/train.py)
- Pipeline step: [dvc.yaml, train_marker](https://github.com/IlyaGusev/rudetox/blob/main/dvc.yaml#L367)
## Eval results
TBA |