project-setup

To set up a CTR model benchmarking environment, this skill clones the BARS and FuxiCTR repositories, creates a Python virtual environment, installs fuxictr>=2.3.7, and prepares a centralized data directory for training experiments.

Linking shared dataset directories for FuxiCTR involves creating or connecting a centralized data folder, then copying dataset and model configs from BARS to FuxiCTR and adjusting config paths to absolute locations before running training.

Yes, you can run DCNv2 CTR training on MovielensLatest_x1 without manual environment drift by using this setup workflow to provision repositories, wire data directories, and prepare model configurations automatically.

The FuxiCTR environment setup optionally applies a NumPy compatibility patch to ensure models run correctly, addressing potential version conflicts after installing fuxictr>=2.3.7 in the Python virtual environment.

The best way to prepare BARS and FuxiCTR workspaces for baseline CTR evaluations is to run an automated provisioning step that clones repositories, installs dependencies, links datasets, and copies model configs for immediate training.