signate-deploy 0.1.7

A CLI tool to automate SIGNATE competition workflows via GitHub Actions

signate-deploy

A CLI tool to automate SIGNATE competition workflows via GitHub Actions.

git push → GitHub Actions → Download data → Train → Submit to SIGNATE

signate-deploy init-repo              # Set up GitHub Actions workflows
signate-deploy setup-token            # Get SIGNATE token & set GitHub Secret
signate-deploy competition-list       # List available competitions
signate-deploy task-list <comp_key>   # Get task_key from competition
signate-deploy file-list <task_key>   # Get file_keys from task
signate-deploy init my-comp \
  --task-key <task_key> \
  --file-key train:<key> \
  --file-key test:<key>               # Create competition directory
signate-deploy submit my-comp \
  --memo "Baseline v1"               # Trigger train & submit
signate-deploy download my-comp      # Trigger data download only

Installation

pip install signate-deploy

Prerequisites: GitHub CLI (gh) must be installed and authenticated.

Quick Start

1. Install signate CLI

pip install signate

The signate CLI is required to fetch your API token and get file keys. It is a separate package from signate-deploy.

2. Set up GitHub Actions

In your GitHub repository root:

python -m signate_deploy init-repo

Creates:

• .github/workflows/signate-submit.yml — full pipeline (download → train → submit)
• .github/workflows/signate-download.yml — data download only
• scripts/refresh_signate_token.py — auto token refresh script

3. Set SIGNATE credentials as GitHub Secrets

Run inside your repository directory:

gh secret set SIGNATE_EMAIL --body [e-mail]
gh secret set SIGNATE_PASSWORD   # prompted securely
gh secret set WANDB_API_KEY      # if using W&B experiment tracking

The generated scripts/refresh_signate_token.py is called at the start of every Actions run to obtain a fresh token automatically — no manual token refresh needed.

Windows (cmd.exe): Use python -m signate_deploy instead of signate-deploy since the Scripts folder may not be on PATH.

4. Get task_key and file_keys

# Find your competition_key from the competition URL:
# https://user.competition.signate.jp/.../detail/?competition=THIS_IS_COMPETITION_KEY

# Get task_key from competition_key
python -m signate_deploy task-list <competition_key>

# Get file_keys from task_key
python -m signate_deploy file-list <task_key>

You can also browse available competitions with python -m signate_deploy competition-list.

5. Create competition directory

python -m signate_deploy init my-comp \
  --task-key abc123def456 \
  --file-key train:5f0e1ebb35af4963 \
  --file-key test:72f23ebe8f004fa0 \
  --file-key sample_submit:ad3502af26b9

Creates:

my-comp/
  signate-config.json   # task_key and file_keys
  train.py              # LightGBM 5-fold CV template
  requirements.txt      # pandas, numpy, scikit-learn, lightgbm

6. Edit train.py and push

# Edit my-comp/train.py (set TARGET column name, add preprocessing, etc.)
git add my-comp/ && git commit -m "Add my-comp baseline" && git push

7. Submit

python -m signate_deploy submit my-comp --memo "Baseline v1"
# → gh workflow run signate-submit.yml is triggered

# Check progress
gh run list --limit 1
gh run view --log

signate-config.json

{
  "task_key": "your_task_key",
  "file_keys": {
    "train": "file_key_for_train_csv",
    "test": "file_key_for_test_csv",
    "sample_submit": "file_key_for_sample_submit_csv"
  }
}

Notes

• ⚠️ Never commit data/ or .signate/ — they are .gitignored by init-repo
• Requires GitHub CLI (gh) to be installed and authenticated
• Works on any OS (Windows/Mac/Linux)

Links

• PyPI
• Article (JP)
• Related: kaggle-notebook-deploy