How anshm42/f26-nuclearchallenge works
This research prototype turns whole NPPAD simulation CSVs into leakage-controlled train, validation, and test splits, then trains calibrated binary leak and optional leak-type models. The main workflow is split runs, summarize early sensor histories, train and select a binary alert threshold on validation data, and evaluate the frozen artifact on held-out runs. A separate conditional branch classifies leak type. Edges reflect the shown CLI calls and README workflow; evaluation outputs and several type-model details are documented but only partly excerpted.
Main entry points: cli.py Model CLI · features.py Split loader · nuclear_dataset_splitter.py Run splitter · model.py Binary metrics
Components 15
Data partitioning
- Run splitter
nuclear_dataset_splitter.py - Split CSV runs
Feature preparation
- Split loader
src/leak_detection/features.py - Run discovery
src/leak_detection/features.py - Run features
src/leak_detection/features.py
Binary leak model
- Calibrated fitting
src/leak_detection/model.py - Threshold selection
src/leak_detection/model.py - Binary metrics
src/leak_detection/model.py - Binary artifact
Leak-type model
- Type fitting
src/leak_detection/model.py - Type metrics
src/leak_detection/model.py - Type artifact
CLI and results
- Model CLI
src/leak_detection/cli.py - Evaluation outputs
Other components
- Researcher
Connections 19
- Researcher → Run splitter: runs split
README.md - Run splitter → Split CSV runs: copies runs
nuclear_dataset_splitter.py - Researcher → Model CLI: runs commands
README.md - Split CSV runs → Model CLI: provides splits
README.md - Model CLI → Split loader: loads splits
src/leak_detection/cli.py - Split loader → Run discovery: discovers runs
src/leak_detection/features.py - Split loader → Run features: extracts features
src/leak_detection/features.py - Model CLI → Calibrated fitting: fits model
src/leak_detection/cli.py - Model CLI → Threshold selection: selects threshold
src/leak_detection/cli.py - Model CLI → Binary metrics: computes metrics
src/leak_detection/cli.py - Model CLI → Binary artifact: saves artifact
src/leak_detection/cli.py - Model CLI → Type fitting: fits type model
src/leak_detection/cli.py - Model CLI → Type metrics: computes type metrics
README.md - Model CLI → Type artifact: saves type artifact
src/leak_detection/cli.py - Binary artifact → Model CLI: loads for evaluation
src/leak_detection/cli.py - Type artifact → Model CLI: loads for evaluation
src/leak_detection/cli.py - Model CLI → Evaluation outputs: writes predictions
src/leak_detection/cli.py - Model CLI → Binary metrics: evaluates held-out runs
src/leak_detection/cli.py - Model CLI → Type metrics: evaluates leak types
README.md