Model Card for Model ID

Model Details

Model Description

  • Developed by: @afwull
  • Model type: Multilanguage
  • Language(s) (NLP): No language left behind languages
  • Finetuned from model: facebook/nllb-200-3.3B

Uses

Tatar text detoxification.

Bias, Risks, and Limitations

Focused on internet at comments. Dataset bias shifts to political comments.

How to Get Started with the Model

Use the code below to get started with the model.

[More Information Needed]

Training Details

Training Data

https://huggingface.co/datasets/afwull/tatar_detox

Training Procedure

SFT Train Loss: Cross-Entropy with label_smoothing=0.1 4 epochs LR: 3e-4

Training Hyperparameters

  • Training regime: bf16

Metrics

Scored 0.37 at PAN 2025. (https://pan.webis.de/clef25/pan25-web/)

Downloads last month
-
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for afwull/NLLB-3.3b-tatar-detox

Adapter
(17)
this model