R + Weights & Biases Integration Example

This project demonstrates how to train a model in R and log metrics to Weights & Biases using a Python bridge.

Structure

• src/R/models/LinearModelTrainer.R: R6 class for model training.
• src/R/utils/WandbLogger.R: R6 class that handles communication with Python.
• src/python/wandb_adapter.py: Python script that interfaces with the W&B SDK.
• main.R: Main entry point.

Prerequisites

1. R Dependencies

Run the install script:

Rscript install.R

2. Python Dependencies

You need wandb installed.

pip install -r requirements.txt

3. Environment Setup

1. Copy .Renviron.example to .Renviron:

   cp .Renviron.example .Renviron

2. Edit .Renviron and add your W&B API Key:

   WANDB_API_KEY=your_actual_api_key
   

   Or ensure you are logged in via CLI:

   wandb login

Usage

Run the main R script:

Rscript main.R

Or source main.R in your RStudio session and run main().

How it works

1. R trains the model and calculates metrics.
2. WandbLogger in R writes these metrics to a temporary JSON file.
3. R calls src/python/wandb_adapter.py via system2().
4. The Python script reads the JSON, logs to W&B, and exits.