mmengine.device.get_max_cuda_memory

mmengine.device.get_max_cuda_memory(device=None)

Returns the maximum GPU memory occupied by tensors in megabytes (MB) for a given device. By default, this returns the peak allocated memory since the beginning of this program.

Parameters:

device (torch.device, optional) – selected device. Returns statistic for the current device, given by current_device(), if device is None. Defaults to None.

Returns:

The maximum GPU memory occupied by tensors in megabytes for a given device.

Return type:

int