unique-func/adialab2026_distilledoracle

How unique-func/adialab2026_distilledoracle works

This repository is a structural-break detection competition artifact: it trains three model packages, combines their predictions, and exposes a Crunch inference entrypoint. The documented training flow runs Anil, Tarik, and Cihat stages before fitting the blend; inference loads the shipped packages and produces predictions. The graph preserves each package’s main modeling approach and the integer-grid detector. Some package internals are unsampled, so those components are represented at their supported source or directory paths rather than expanded speculatively.

Main entry points: main.py Root pipeline · anil/main.py Anil pipeline · cihat/main.py Cihat entry · run.py Cihat training

Components 16

Submission entry

Root pipeline main.py
Grid detector precision_grid.py

Anil models

Anil pipeline anil/main.py
Streaming features anil/features/feature_window_scan.py
RealMLP model anil/models/pointwise_realmlp/train.py

Tarik models

Tarik entry tarik/main.py
Tarik runtime tarik/sbrt/deployment/current.py
Retained features tarik/sbrt/deployment/retention_graph.py

Cihat models

Cihat entry cihat/main.py
Cihat training cihat/run.py
Cihat predictors cihat/submission
RMV features cihat/trainlib/rmv.py
Volatility model cihat/trainlib/volatility/model.py

Ensemble output

Prediction blend blend.py
Blend fitting blend_fit.py

Other components

Crunch runner

Connections 16

Crunch runner → Root pipeline: invokes README.md
Root pipeline → Anil pipeline: trains and infers main.py
Root pipeline → Tarik entry: trains and infers main.py
Root pipeline → Cihat entry: trains and infers main.py
Root pipeline → Grid detector: uses detector main.py
Root pipeline → Prediction blend: combines members main.py
Anil pipeline → Streaming features: builds features anil/main.py
Anil pipeline → RealMLP model: trains model anil/main.py
Tarik entry → Tarik runtime: dispatches training tarik/main.py
Tarik runtime → Retained features: uses feature graph tarik/sbrt/deployment/current.py
Cihat entry → Cihat training: runs training cihat/main.py
Cihat training → RMV features: fits RMV cihat/trainlib/rmv.py
Cihat training → Volatility model: trains volatility
Cihat entry → Cihat predictors: uses predictors cihat/main.py
Blend fitting → Prediction blend: fits weights README.md
Root pipeline → Blend fitting: fits ensemble README.md

Drawn by GitDiagram on November 15, 2026 from the repository's file tree, README and source files. Read as Markdown ·