Roguebert Toxicity 85K

🚀 Toxicity Classification Model

A fine - tuned model based on RoBERTa - base for detecting comment toxicity, achieving high accuracy on the evaluation set.

🚀 Quick Start

This model is a fine - tuned version of [roberta - base](https://huggingface.co/roberta - base) on the dirtycomputer/Toxic_Comment_Classification_Challenge dataset. It achieves the following results on the evaluation set:

• Loss: 0.0511
• Accuracy: 0.9812

✨ Features

• Fine - tuned roberta - base model for detecting toxicity in comments.
• Categorizes a comment into different toxicity types, such as "toxic," "obscene," "insult," and "threat."

📚 Documentation

Intended Uses & Limitations

Intended Uses

• Content Moderation: Automatically flagging or removing toxic comments on social media platforms, forums, and customer support.
• Toxicity Detection: Classifying comments based on their toxicity level, such as harmful language or insults.

Limitations

• False Negatives: May not always catch subtle toxic behavior.
• Limited Language Support: Currently, the model is trained on English - only data.
• Context Sensitivity: May struggle with ambiguous language or sarcasm.

Training and Evaluation Data

This model was fine - tuned using the dirtycomputer/Toxic_Comment_Classification_Challenge dataset, which contains labeled comments for toxicity classification.

Training Procedure

Training Hyperparameters

The following hyperparameters were used during training:

• learning_rate: 5e - 05
• train_batch_size: 8
• eval_batch_size: 16
• seed: 42
• optimizer: Use adamw_torch with betas=(0.9, 0.999) and epsilon = 1e - 08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• num_epochs: 3

Training Results

Framework Versions

• Transformers 4.47.1
• Pytorch 2.5.1+cu121
• Datasets 3.2.0
• Tokenizers 0.21.0

📄 License

This project is licensed under the MIT license.