vllm-sr/mmbert-intent-classifier-merged

mmBERT Intent Classifier (Merged for Rust)

This is a merged mmBERT model for intent classification, optimized for Rust inference using the candle framework.

Model Details

• —Base Model: jhu-clsp/mmBERT-base
• —Task: 14-class MMLU-Pro category classification
• —Languages: Multilingual (1800+ languages)
• —Training: LoRA fine-tuned then merged with base model
• —Inference: Optimized for Rust candle-binding

Usage with Rust (candle-binding)

use candle_semantic_router::classifiers::lora::IntentLoRAClassifier;

let classifier = IntentLoRAClassifier::new("path/to/model", true)?;
let result = classifier.classify_intent("What is photosynthesis?")?;
println!("Intent: {}, Confidence: {:.1}%", result.intent, result.confidence * 100.0);

Classes

Training Configuration

• —LoRA Rank: 32
• —LoRA Alpha: 64
• —Epochs: 10
• —Batch Size: 64
• —Learning Rate: 2e-5

License

Apache 2.0