lora_jailbreak_classifier_bert-base-uncased_model

lora_jailbreak_classifier_bert-base-uncased_model

Model Description

This is a LoRA (Low-Rank Adaptation) fine-tuned model based on bert-base-uncased for Multi-task classification.

This model is part of the semantic-router project and is optimized for use with the Candle framework in Rust.

Model Details

• Base Model: bert-base-uncased
• Task: Classification Classification
• Framework: Candle (Rust)
• Model Size: ~418MB
• LoRA Rank: 16
• LoRA Alpha: 16
• Target Modules: attention.self.query, attention.self.value, attention.output.dense, intermediate.dense, output.dense

Usage

With semantic-router (Recommended)

from semantic_router import SemanticRouter

# The model will be automatically downloaded and used
router = SemanticRouter()
results = router.classify_batch(["Your text here"])

With Candle (Rust)

use candle_core::{Device, Tensor};
use candle_transformers::models::bert::BertModel;

// Load the model using Candle
let device = Device::Cpu;
let model = BertModel::load(&device, &config, &weights)?;

Training Details

This model was fine-tuned using LoRA (Low-Rank Adaptation) technique:

• Rank: 16
• Alpha: 16
• Dropout: 0.1
• Target Modules: attention.self.query, attention.self.value, attention.output.dense, intermediate.dense, output.dense

Performance

Multi-task classification

For detailed performance metrics, see the training results.

Files

• model.safetensors: LoRA adapter weights
• config.json: Model configuration
• lora_config.json: LoRA-specific configuration
• tokenizer.json: Tokenizer configuration
• label_mapping.json: Label mappings for classification

Citation

If you use this model, please cite:

@misc{semantic-router-lora,
  title={LoRA Fine-tuned Models for Semantic Router},
  author={Semantic Router Team},
  year={2025},
  url={https://github.com/vllm-project/semantic-router}
}

License

Apache 2.0