yandex-research/rtdl-revisiting-models

How yandex-research/rtdl-revisiting-models works

This repository is the official research implementation for comparing deep-learning architectures on tabular data and reproducing the paper’s benchmark results. The principal workflow is to load and preprocess a dataset, train a configured model, record metrics, tune hyperparameters, and optionally ensemble results; a separate analysis compares FT-Transformer and ResNet using synthetic data. The graph preserves the major model families and experiment stages. Source excerpts show only parts of the runtime, so links between model runners and shared utilities rely on the README’s description of lib as their common tools.

Main entry points: data.py Dataset processing · resnet.py MLP and ResNet · analysis_ft_transformer_vs_resnet.py Architecture comparison · tune.py Hyperparameter tuning

Components 22

Experiments

Hyperparameter tuning bin/tune.py
Ensembling bin/ensemble.py

Model runners

MLP and ResNet bin/resnet.py
FT-Transformer bin/ft_transformer.py
AutoInt bin/autoint.py
DCN v2 bin/dcn2.py
GrowNet bin/grownet.py
NODE bin/node.py
SNN bin/snn.py
TabNet bin/tabnet.py
Tree models bin/catboost_.py

Shared runtime

Dataset processing lib/data.py
Training utilities lib/deep.py
Metrics lib/metrics.py
NODE architecture lib/node/arch.py
Differentiable trees lib/node/odst.py
NODE NN utilities lib/node/nn_utils.py

Analysis

Architecture comparison bin/analysis_ft_transformer_vs_resnet.py
Synthetic data lib/synthetic_data.py

Other components

Researcher
Tabular dataset
Experiment outputs

Connections 26

Researcher → Hyperparameter tuning: configures runs README.md
Hyperparameter tuning → MLP and ResNet: tunes model README.md
Researcher → MLP and ResNet: runs experiments README.md
Tabular dataset → Dataset processing: provides inputs README.md
MLP and ResNet → Dataset processing: uses common tools README.md
FT-Transformer → Dataset processing: uses common tools bin/ft_transformer.py
AutoInt → Dataset processing: uses common tools bin/autoint.py
DCN v2 → Dataset processing: uses common tools bin/dcn2.py
GrowNet → Dataset processing: uses common tools bin/grownet.py
NODE → Dataset processing: uses common tools bin/node.py
SNN → Dataset processing: uses common tools bin/snn.py
TabNet → Dataset processing: processes dataset bin/tabnet.py
Tree models → Dataset processing: uses common tools bin/catboost_.py
MLP and ResNet → Training utilities: uses training tools README.md
FT-Transformer → Training utilities: uses training tools README.md
MLP and ResNet → Metrics: reports metrics
MLP and ResNet → Experiment outputs: writes results README.md
Ensembling → Experiment outputs: reads run results README.md
Ensembling → Experiment outputs: writes ensembles README.md
Hyperparameter tuning → Experiment outputs: records configurations README.md
NODE → NODE architecture: uses architecture bin/node.py
NODE architecture → Differentiable trees: builds layers lib/node/arch.py
Differentiable trees → NODE NN utilities: uses operators lib/node/odst.py
Architecture comparison → FT-Transformer: compares model bin/analysis_ft_transformer_vs_resnet.py
Architecture comparison → MLP and ResNet: compares model bin/analysis_ft_transformer_vs_resnet.py
Architecture comparison → Synthetic data: uses generated data bin/analysis_ft_transformer_vs_resnet.py

Drawn by GitDiagram on November 12, 2026 from the repository's file tree, README and source files. Read as Markdown ·