subject_classifier_19_subjects

DistilBERT Subject Classifier

A lightweight, single‐label text‐classification model built on DistilBERT to assign one of 19 high‑level academic or professional subjects to an input text.

---

Model Overview

• Architecture: DistilBertForSequenceClassification
• Hidden Size: 768
• Intermediate (FFN) Size: 3072
• Heads / Layers: 12 heads × 6 layers
• Dropout:
  • Attention dropout: 0.1
  • Sequence‐classification dropout: 0.2
• Activation: GELU
• Max Sequence Length: 512 tokens

---

Label Mapping

---

Configuration Highlights

{
  "model_type": "distilbert",
  "initializer_range": 0.02,
  "problem_type": "single_label_classification",
  "torch_dtype": "float32",
  "transformers_version": "4.49.0"
}