Mental Health Text Classifier (QLoRA, BERT) This is a small and efficient model built with bert-base-uncased and fine-tuned under Quantized Low Rank Adaptation. It classifies mental health-related text into categories like Anxiety, Depression, etc.

DistilBert Info

Type: Natural Language text sentiment classifier using Google BERT + QLoRA adapters Purpose: Detect mental health sentiments througgh short text Uses: Social media, forums, research, screening

Originally prototyped for community health project as a Fairfax County Public Youth Health Ambassador

Not for Medical diagnosis or treatment

Details Created by: Bhavesh Adivi Data: "Kaggle mental health sentiment dataset" Hardware: Trained on Nvidia Tesla 4 GPU - Access provided by Google Colab License: Apache 2.0

Usage & Inference: Can be used to (experimentally) classify user posts or comments for mental health signs Use in apps for (experimental) mental health monitoring or (experimental only) early alerts Can be further fine-tuned for better results or distinct cohorts

Important Notes to consider: This is NOT meant for immediate diagnosis or replacing health professionals Bias risk included: Potential data biases may affect results from datasets Some Limitations might include missing indirect or subtle cues such as sarcasm or critical context Make sure to always involve human judgment and privacy protections

Model is hosted on Hugging Face for inference: https://huggingface.co/spaces/bhaveshadivi/distilMind