Rate this Page
★★★★★
torch.accelerator.memory.memory_allocated#
- torch.accelerator.memory.memory_allocated(device_index=None,/)[source]#
Return the currentaccelerator device memory occupied by tensorsin bytes for a given device index.
- Parameters
device_index (
torch.device, str, int, optional) – the index of the device to target.If not given, usetorch.accelerator.current_device_index()by default.If atorch.deviceor str is provided, its type must match the currentaccelerator device type.- Return type
On this page
