deberta_mental / README.md
elishaw's picture
README.md
6cf92cb verified
metadata
tags:
  - text-classification
  - mental-health
  - deberta-v3
  - pytorch
  - transformers
  - sentiment-analysis
  - healthcare
language:
  - en
license: mit
datasets:
  - AIMH/SWMH
metrics:
  - accuracy
  - f1
pipeline_tag: text-classification

DeBERTa Mental Health Classification Model

A fine-tuned DeBERTa v3 small model for detecting mental health conditions from text.

Model Description

This model is based on microsoft/deberta-v3-small and has been fine-tuned to classify text into 8 mental health categories.

Training Data

This model was trained on the following datasets:

Labels

The model can classify text into the following categories:

ID Label Description
0 Normal No mental health concerns detected
1 Offmychest General venting/sharing
2 Depression Depression-related content
3 Anxiety Anxiety-related content
4 Stress Stress-related content
5 Bipolar Bipolar disorder-related content
6 Personality disorder Personality disorder-related content
7 Suicidal Suicidal ideation (⚠️ requires immediate attention)

Usage

from transformers import AutoTokenizer, AutoModelForSequenceClassification
import torch

# Load model and tokenizer
model_path = "deberta-illness"
tokenizer = AutoTokenizer.from_pretrained(model_path)
model = AutoModelForSequenceClassification.from_pretrained(model_path)

# Example text
text = "I've been feeling down lately and can't seem to enjoy anything anymore."

# Tokenize and predict
inputs = tokenizer(text, return_tensors="pt", truncation=True, max_length=512)
outputs = model(**inputs)
predictions = torch.nn.functional.softmax(outputs.logits, dim=-1)

# Get predicted label
predicted_class = torch.argmax(predictions, dim=-1).item()
confidence = predictions[0][predicted_class].item()

print(f"Predicted: {model.config.id2label[str(predicted_class)]}")
print(f"Confidence: {confidence:.2%}")

Model Architecture

  • Base Model: microsoft/deberta-v3-small
  • Hidden Size: 768
  • Attention Heads: 12
  • Hidden Layers: 6
  • Max Sequence Length: 512 tokens
  • Vocabulary Size: 128,100

License

Please refer to the original microsoft/deberta-v3-small license and any additional licensing terms from the fine-tuning dataset.