hrshlgunjal/emotion-classifier-roberta-base

Emotion Classification Model

This model classifies text into 5 emotion categories: anger, fear, joy, sadness, and surprise.

Model Description

• Base Model: roberta-base
• Task: Multi-label text classification
• Labels: anger, fear, joy, sadness, surprise
• Training Strategy: 5-Fold Cross-Validation
• Framework: PyTorch + Transformers

Performance

Overall Metrics

• Macro F1: N/A
• Cross-Validation: 0.7684 +/- 0.0099

Per-Label Performance

N/A

Optimized Thresholds

N/A

Usage

from transformers import AutoTokenizer, AutoModelForSequenceClassification
import torch
import numpy as np

# Load model and tokenizer
model = AutoModelForSequenceClassification.from_pretrained("hrshlgunjal/emotion-classifier-roberta-base")
tokenizer = AutoTokenizer.from_pretrained("hrshlgunjal/emotion-classifier-roberta-base")

# Optimized thresholds (use these for best results)
thresholds = np.array([0.5, 0.5, 0.5, 0.5, 0.5])
labels = ['anger', 'fear', 'joy', 'sadness', 'surprise']

# Predict emotions
def predict_emotions(text):
    inputs = tokenizer(text, return_tensors="pt", truncation=True, max_length=128)
    with torch.no_grad():
        outputs = model(**inputs)
    probs = torch.sigmoid(outputs.logits).cpu().numpy()[0]
    predictions = (probs >= thresholds).astype(int)
    return {label: (pred, prob) for label, pred, prob in zip(labels, predictions, probs)}

# Example
text = "I am so excited about this amazing opportunity!"
result = predict_emotions(text)
print(result)

Training Details

• Optimizer: AdamW with differential weight decay
• Learning Rate: 2e-05
• Batch Size: 32
• Epochs: 3
• Max Sequence Length: 128
• Warmup Ratio: 0.1
• Weight Decay: 0.01
• Mixed Precision: Enabled (FP16)
• Gradient Clipping: 1.0

Training Infrastructure

• Device: GPU
• Training Time: ~225 minutes (approximate)
• Framework Versions:
  • PyTorch: 2.6.0+cu124
  • Transformers: 4.53.3

Model Card Authors

hrshlgunjal

Model Card Contact

For questions or feedback, please open an issue in the model repository.