torch.cuda.memory.memory_allocated#

torch.cuda.memory.memory_allocated(device=None)[source]#

Return the current GPU memory occupied by tensors in bytes for a given device.

Parameters

device (torch.device orint,optional) – selected device. Returnsstatistic for the current device, given bycurrent_device(),ifdevice isNone (default).

Return type

int

Note

This is likely less than the amount shown innvidia-smi since someunused memory can be held by the caching allocator and some contextneeds to be created on GPU. SeeMemory management for moredetails about GPU memory management.