o Zh@shddlZddlmZmZddlmZmZddefdd Z dd d Z d d Z ddZ dddZ ddZdS)N) tree_flattentree_unflatten)add_step_closurerun_step_closuresFdevicecCstjjj|g|dtdS)aVTriggers a mark step, which amounts to - collecting a group of 'live' lazy tensors to index into the compilation cache (lowering/compiling their IR graphs if not cached) - kicking off execution of the compiled function - (optionally, wait=True) waiting for cpu-side execution to complete (does not sync the accelerator) )waitN)torch_C_lazyZ _mark_stepr)rr r C/var/www/auris/lib/python3.10/site-packages/torch/_lazy/__init__.py mark_step s rcCs |durg}tjjj|ddS)zWaits for all the async operations on the given devices to complete. Args: devices (string..., optional): The devices whose async ops need to be waited for. If empty, all the local devices will be waited for. Ndevices)r r r Z_wait_device_opsrr r rwait_device_opssrcCstjj||dS)z Sync the list of lazy tensors so there IR get lowered for the activate backend and the compiled computation graph get cached. N)r r r Z _sync_multi)tensorsrr r r sync_multi!srcCstjj|S)z7Return a unique id of the lazy tensor maintained by LTC)r r r Z_get_tensor_id)Ztensorr r r get_tensor_id)srcCs4|pdg}t|\}}t||tdd|D|S)NZlazycSsg|]}|dqS)cpu)to).0tr r r 3szto_cpu..)rrr)rrZ flattenedspecr r rto_cpu.s   rcOs tjt|g|Ri|dSN)r saver)rargskwargsr r rr6s r)rFr)Ztorch._C._lazyr Ztorch.utils._pytreerrclosurerrstrrrrrrrr r r rs