Repository navigation
Configuration
Sudhanshu edited this page Nov 11, 2026
·
1 revision
The source defines its settings as Python constants and TrainingArguments, not environment variables or a YAML configuration file.
| Setting | Value |
|---|---|
SEED |
42 |
MODEL_NAME |
distilbert-base-uncased |
LABEL_NAMES |
billing, technical, account, feature_request |
| Split | 80/20 stratified, 24/6 |
| Maximum length | 128 tokens |
| Full fine-tune | 4 epochs, batch 8, LR 2e-5 |
| Frozen experiment | 6 epochs, batch 8, LR 1e-3 |
| Runtime | CPU or CUDA; BF16/FP16 selected if available |
The repository does not support parameterizing these values via command line without code modifications.
Getting started
Architecture
- System Context
- High-Level Design
- Low-Level Design
- Model Components
- Training Sequence
- Data & Labels
- Tokenization
Training & results
Engineering reference