### Install dependencies

Source: https://github.com/yandex-research/tabpack/blob/main/README.md

Install the project dependencies using uv.

```shell
uv sync
```

--------------------------------

### Run quick test demo

Source: https://github.com/yandex-research/tabpack/blob/main/README.md

Execute the demo script to verify the environment setup.

```shell
uv run -m lib.examples.demo experiments/examples/demo --force
```

--------------------------------

### Create experiment directory

Source: https://github.com/yandex-research/tabpack/blob/main/README.md

Create a new subdirectory for reproducing experiments.

```shell
mkdir -p experiments/reproduce/tabpack
```

--------------------------------

### Generate experiment configs

Source: https://github.com/yandex-research/tabpack/blob/main/README.md

Run the factory script to generate experiment configurations and commands.

```shell
uv run experiments/reproduce/tabpack/make.py
```

--------------------------------

### Run experiments

Source: https://github.com/yandex-research/tabpack/blob/main/README.md

Execute the generated shell script to run the experiments.

```shell
sh experiments/reproduce/tabpack/commands.sh
```

--------------------------------

### Clone the repository

Source: https://github.com/yandex-research/tabpack/blob/main/README.md

Clone the TabPack repository and navigate into the directory.

```shell
git clone https://github.com/yandex-research/tabpack
cd tabpack
```

--------------------------------

### Download datasets

Source: https://github.com/yandex-research/tabpack/blob/main/README.md

Download and extract the primary datasets.

```shell
wget https://huggingface.co/datasets/Yura52/tabpack-data/resolve/main/tabpack-data.tar.gz
tar -xvf tabpack-data.tar.gz
```

--------------------------------

### View generated commands

Source: https://github.com/yandex-research/tabpack/blob/main/README.md

Display the contents of the generated command file.

```shell
cat experiments/reproduce/tabpack/commands.sh
```

--------------------------------

### Check generated files

Source: https://github.com/yandex-research/tabpack/blob/main/README.md

List the contents of the reproduction directory.

```shell
ls experiments/reproduce/tabpack
```

--------------------------------

### Experiment directory structure

Source: https://github.com/yandex-research/tabpack/blob/main/README.md

The directory layout for experiments, including main runs, evaluation runs, and configuration generation scripts.

```text
experiments/
  <model>/
    <dataset>/
      main/  # The main run.
      eval-*-ensembles/
        greedy/
          evaluation/  # The evaluation run.
    make.py  # A script for generating experiment configs and commands.
```

--------------------------------

### Run code quality checks

Source: https://github.com/yandex-research/tabpack/blob/main/README.md

Run linting and type checking tools.

```shell
uv run ruff check
uv run ruff format --check
uv run ty check
```

--------------------------------

### Copy experiment script

Source: https://github.com/yandex-research/tabpack/blob/main/README.md

Copy the experiment factory script to the reproduction directory.

```shell
cp experiments/tabpack/make.py experiments/reproduce/tabpack
```

--------------------------------

### Edit commands file

Source: https://github.com/yandex-research/tabpack/blob/main/README.md

Open the commands file to filter experiments if necessary.

```shell
$EDITOR experiments/reproduce/tabpack/commands.sh
```

--------------------------------

### Summarize evaluation metrics

Source: https://github.com/yandex-research/tabpack/blob/main/README.md

This Python script loads experiment reports from the evaluation directory and calculates the mean and standard deviation of test scores across ensemble runs.

```python
import statistics

import lib.datasets
import lib.experiment

model = 'tabpack-cosine'
for dataset in lib.datasets.MAIN_DATASETS:
    exp = f'experiments/{model}/{dataset}/eval-online-ensembles/greedy/evaluation'
    report = lib.experiment.load_report(exp)
    scores = [
        x['report']['online_ensembles']['greedy']['report']['metrics']['test']['score']
        for x in report['experiments']
    ]
    print(
        f'{dataset:<25}'
        f'  {abs(statistics.mean(scores)):>11.4f}'
        f' +- {abs(statistics.stdev(scores)):>7.4f}'
    )
```

--------------------------------

### BibTeX citation for TabPack

Source: https://github.com/yandex-research/tabpack/blob/main/README.md

Use this BibTeX entry to cite the TabPack paper in your research.

```bibtex
@inproceedings{gorishniy2026tabpack,
    title={{TabPack: Efficient Hyperparameter Ensembles for Tabular Deep Learning}},
    author={Yury Gorishniy and Akim Kotelnikov and Ivan Rubachev and Artem Babenko},
    booktitle={ICML},
    year={2026},
}
```

=== COMPLETE CONTENT === This response contains all available snippets from this library. No additional content exists. Do not make further requests.