SCCMR-ANGST: Multi-Label Mental Health Classification

Fine-tuned from SCCMR-MDA for multi-label Depression and Anxiety detection.

Model Description

Predicts Depression and/or Anxiety from text posts.

Outputs:

• Depression only: [1, 0]
• Anxiety only: [0, 1]
• Both: [1, 1]
• Neither: [0, 0]

Training Data

• Dataset: ANGST
• Train samples: 4830
• Val samples: 1035
• Test samples: 1035

Usage

import torch
from transformers import BertTokenizer

# Load model and tokenizer
checkpoint = torch.load("best_multilabel_model.pth")
model.load_state_dict(checkpoint['model_state_dict'])
tokenizer = BertTokenizer.from_pretrained("alfiyahqthz/sccmr-angst")

# Predict
text = "I feel depressed and anxious"
inputs = tokenizer(text, return_tensors="pt", padding=True, truncation=True)
logits = model(inputs['input_ids'], inputs['attention_mask'])
probs = torch.sigmoid(logits)
predictions = (probs > 0.5).long()

License

MIT