loke007/bugflow-severity-classifier

BugFlow Severity Classifier

Fine-tuned CodeBERT model for classifying bug report severity levels.

Labels

• Low: Minor issues, cosmetic changes
• Medium: Standard bugs affecting some functionality
• High: Important bugs affecting major functionality
• Critical: System crashes, data loss, security issues

Usage

from transformers import RobertaTokenizer, RobertaForSequenceClassification
import torch

model = RobertaForSequenceClassification.from_pretrained("YOUR_USERNAME/bugflow-severity-classifier")
tokenizer = RobertaTokenizer.from_pretrained("YOUR_USERNAME/bugflow-severity-classifier")

text = "Application crashes when clicking login button"
inputs = tokenizer(text, return_tensors="pt", truncation=True, padding=True)
outputs = model(**inputs)
probs = torch.softmax(outputs.logits, dim=1)
labels = ['low', 'medium', 'high', 'critical']
predicted = labels[torch.argmax(probs).item()]
print(f"Severity: {predicted}")

Training

• Base model: microsoft/codebert-base
• Dataset: Custom GitHub issues dataset + domain-specific bugs
• Fine-tuned using Hugging Face Transformers