dual-folder-workflow
Installation
SKILL.md
Dual-Folder Workflow - Research Notes
Experiment Overview
| Item | Details |
|---|---|
| Date | 2025-12-18 |
| Goal | Separate development from production when using private repo with Colab training |
| Environment | Windows, Git, Google Colab, Private GitHub repo |
| Status | Success |
Context
When a repository is private, Google Colab cannot directly git clone it. The typical workflow involves:
- Zip the local repo
- Upload to Google Drive
- Unzip in Colab and train
The problem: If you're actively developing new features, you risk accidentally uploading untested code for production training. Models trained with buggy code could then be used for live trading.