Base models

An allowed model is available for use in this deployment. The deployment’s hardware and training engines must support it. You can’t add models to this list.

A selected model is an allowed model that you have prepared for training, evaluation, or serving. Selecting a model registers it. If the deployment uses a shared cache, selection also downloads the weights in advance. Initially, no models are selected.

List the selected models:

To include every allowed model, add --all:

The repository identifies each model and is the value that other base model commands accept. The output also shows the pinned commit, parameter count, context length, and training methods that the hardware supports. STATUS shows the model’s selection state. A selected and prepared model is ready. A selected model whose weights haven’t arrived is preparing. An allowed model that nobody has selected is not-selected.

The allowed list includes only models that the training engines and hardware support. If the list doesn’t contain a model that you need, ask your operator to add it to the deployment.

To see what a run over the model can request, get the model. Under Runs this deployment admits, the output lists every admitted combination of training method, precision, and GPU count. It also shows the memory of each GPU, the host memory of the node that holds them, and the maximum sequence length:

If a run requests an unsupported configuration, the trainer rejects it during submission before using a GPU. For the validation rules, see Base models.

Before you ask an operator to allow a model, check whether the deployment can train it. The candidates command lists repositories from a Hugging Face account and evaluates each model against the deployment’s engines and hardware:

--match filters repository names by using a glob pattern. In the pattern, * matches any sequence of characters. Omit --match to list all repositories from the account.

In the output, VERDICT is admissible if the deployment can train the model. Otherwise, the column explains why the model isn’t supported. Possible reasons include an unsupported architecture, a model that is too large for the hardware, or a model that the trainer can’t adapt.

The command reads data from Hugging Face but doesn’t modify the deployment. It doesn’t allow or select any models.

To generate catalog entries for the admissible models, an operator can add --catalog:

The command returns immediately while preparation continues in the background. Add --wait to wait until the model is ready. The wait ends after 30 minutes. To wait longer, use --timeout.

While the model is being prepared, you can start a run or evaluation, or create an endpoint. The trainer accepts the request and starts the work when the model is ready. In a script, use --wait to prepare the model before the script continues.

To check the preparation progress, run base-models get.

The model remains allowed, and you can select it again. The trainer retains its registration because models trained from it still refer to it.

Deselecting does not stop an endpoint that is already serving the model. To stop one, see Endpoints.