public static supportedQuantizations(repository as String, device as Device, revision as String = "main") as ArrayList<ModelQuantization>
Lists published weight formats executable by this provider on the selected device. CPU supports all published variants implemented by this provider. Reserved NPU and unavailable GPUs return an empty list. AUTO is not a checkpoint variant and is never included. This is a backend-format capability query, not a memory-fit guarantee: context, activations and other processes also consume device memory, and the model architecture is validated when its descriptors are loaded.
| Parameter | Description |
|---|---|
repository | Hugging Face owner/name. |
device | CPU, concrete GPU or category/automatic selector. |
revision | Branch, tag or immutable commit. |
Published variants supported by this execution backend.
print(LocalLLMProvider.supportedQuantizations("bartowski/Qwen_Qwen3.5-9B-GGUF", Device.gpu[0]))