torch.mtia.record_memory_history#
- torch.mtia.record_memory_history(enabled='all', stacks='python', max_entries=0)[source]#
Enable/Disable the memory profiler on MTIA allocator
- Parameters
enabled (all or state, optional) – statistics for the current device, given by current_device(), if device is None (default).
stacks ("python" or "cpp", optional) –
max_entries (int, optional) –