Repository navigation
Source Code Map
Sudhanshu edited this page Nov 11, 2026
·
1 revision
| Area | Actual code |
|---|---|
| Environment | Imports, SEED, CUDA/BF16/FP16 detection |
| Dataset |
RAW_DATA, LABEL_NAMES, label mappings and stratified split |
| Preprocessing |
tokenize_function, DatasetDict.map, DataCollatorWithPadding
|
| Training helpers |
compute_metrics, create_training_arguments, create_trainer
|
| Full experiment |
AutoModelForSequenceClassification, Trainer.train, evaluation and save |
| Inference |
classify_query plus five example support tickets |
| Frozen experiment | Second pretrained model, requires_grad=False on encoder, separate Trainer |
The original source script and notebook are intentionally left unchanged in the complete package. original/README.md retains the previously uploaded README for provenance.
Editable .mmd files live under docs/mermaid in the code repository. The relevant code blocks are embedded in this Wiki's HLD, LLD, and topic pages so GitHub renders them without external asset hosting.
Getting started
Architecture
- System Context
- High-Level Design
- Low-Level Design
- Model Components
- Training Sequence
- Data & Labels
- Tokenization
Training & results
Engineering reference