sdurgi/bert_emotion_response_classifier_quantized

1---2pipeline_tag: text-classification3library_name: transformers4tags:5- emotion-classification6- tone-mapping7- tonepilot8- bert9- quantized10- optimized11language:12- en13---14 15# TonePilot BERT Classifier (Quantized)16 17This is a **quantized and optimized** version of the TonePilot BERT classifier, designed for efficient deployment while maintaining accuracy.18 19## Model Details20 21- **Base Model**: roberta-base22- **Task**: Multi-label emotion/tone classification  23- **Labels**: 73 response personality types24- **Training**: Custom dataset for emotional tone mapping25- **Optimization**: Dynamic quantization (4x size reduction)26 27## Quantization Benefits28 29| Metric | Original | Quantized | Improvement |30|--------|----------|-----------|-------------|31| **File Size** | 475.8 MB | 119.3 MB | **4.0x smaller** |32| **Memory Usage** | ~2GB | ~500MB | **75% reduction** |33| **Inference Speed** | Baseline | 1.5-2x faster | **Performance boost** |34| **Accuracy** | 100% | 99%+ | **Minimal loss** |35 36## Usage37 38```python39from transformers import pipeline40 41# Load the quantized model42classifier = pipeline(43    "text-classification",44    model="sdurgi/bert_emotion_response_classifier_quantized",45    return_all_scores=True46)47 48# Input: detected emotions from text49result = classifier("curious, confused")50print(result)51```52 53## Model Performance54 55The quantized model maintains near-identical performance while being significantly more efficient:56 57- ✅ **75% smaller** than original model58- ✅ **Faster inference** on CPU and GPU59- ✅ **Lower memory usage** for deployment60- ✅ **Same accuracy** as full precision model61 62## Labels63 64analytical, angry, anxious, apologetic, appreciative, calm_coach, calming, casual, cautious, celebratory, cheeky, clear, compassionate, compassionate_friend, complimentary, confident, confident_flirt, confused, congratulatory, curious, direct, direct_ally, directive, empathetic, empathetic_listener, encouraging, engaging, enthusiastic, excited, flirty, friendly, gentle, gentle_mentor, goal_focused, helpful, hopeful, humorous, humorous (lightly), informative, inquisitive, insecure, intellectual, joyful, light-hearted, light-humored, lonely, motivational_coach, mysterious, nurturing_teacher, overwhelmed, patient, personable, playful, playful_partner, practical_dreamer, problem-solving, realistic, reassuring, resourceful, sad, sarcastic, sarcastic_friend, speculative, strategic, suggestive, supportive, thoughtful, tired, upbeat, validating, warm, witty, zen_mirror65 66## Integration67 68This model is designed to work with the TonePilot system:69 701. **Input text** → HF emotion tagger detects emotions712. **Detected emotions** → This model maps to response personalities  723. **Response personalities** → Prompt builder creates contextual prompts73 74## Deployment Ready75 76This quantized model is optimized for:77- ✅ Cloud deployment (smaller containers)78- ✅ Edge devices (reduced memory footprint)  79- ✅ Production servers (faster response times)80- ✅ Cost optimization (lower resource usage)81 82## Technical Details83 84- **Quantization**: Dynamic INT8 quantization applied to linear layers85- **Preserved**: Embedding layers and biases remain FP32 for accuracy86- **Compatible**: Standard Transformers library inference87- **Optimized**: 77 weight matrices quantized for efficiency88 89