Datadog Toto 2.0 Local Smoke Test

This folder contains a small script that downloads public daily temperature data, loads a Toto 2.0 checkpoint from Hugging Face, forecasts the last 96 days, and writes a CSV plus plot under outputs/.

Datadog's current Toto 2.0 quick start installs the inference code from:

pip install "toto-2 @ git+https://github.com/DataDog/toto.git#subdirectory=toto2"

This project uses the GitHub ZIP URL in requirements.txt instead, because this computer's git command is not currently behaving from the OneDrive workspace.

Run

Use the Codex-bundled Python if python is not on your PATH:

& 'C:\Users\gmeyp\.cache\codex-runtimes\codex-primary-runtime\dependencies\python\python.exe' -m venv .venv
.\.venv\Scripts\python.exe -m pip install -r requirements.txt
.\.venv\Scripts\python.exe run_toto2_demo.py

The default model is Datadog/Toto-2.0-4m so the first test is small. To try the 22m model:

.\.venv\Scripts\python.exe run_toto2_demo.py --model Datadog/Toto-2.0-22m

Outputs

• daily-min-temperatures.csv: cached public data pulled from the internet.
• outputs/<model>_temperature_forecast.csv: actual values and Toto quantiles.
• outputs/<model>_temperature_forecast.png: context, actuals, median forecast, and p10-p90 uncertainty band.