Command-line model workflows
Discover and download upstream checkpoints, or invoke a verified external deployment from an automation workflow.
Discover the checkpoint
Use the Hugging Face Hub and its official CLI to inspect and download the exact repository revision you intend to evaluate. Read the model card and license before selecting files. An ONNX tag does not tell you whether the graph includes a forecasting decoder or only an encoder.
Run the actual implementation
Follow upstream Python examples for local model execution, or the chosen provider's CLI and HTTP instructions for remote inference. Preserve the sampling cadence, cutoff, horizon, and model revision with results. The former TSFM operation-discovery and API-key commands do not describe a current service.
Automate with explicit failure states
Pin dependencies and model revisions, bound payload sizes, log throughput for batch jobs, and distinguish download, authentication, shape, and inference failures. Do not silently skip scheduled forecasts or substitute another model under the original checkpoint name.