profile

→Iterator[Profiler]

profile(activities: list[str] | None = None, with_memory: bool = True)

source

Context manager that records ops executed within the with block.

The canonical entry point for ad-hoc profiling: enter the context, run any Lucid computation, exit, then inspect prof.events() or prof.key_averages(). The activities parameter is accepted for parity with the conventional profiling API but is not interpreted.

Parameters

activitieslist of str= None

Currently ignored; reserved for future per-device filtering.

with_memorybool= True

Include memory statistics in the recorded trace. Default True.

Notes

Internally instantiates a Profiler and routes its lifecycle through __enter__ / __exit__. Equivalent to::

prof = Profiler(with_memory=with_memory)
prof.start()
try: ...
finally: prof.stop()

Examples

>>> import lucid
>>> with lucid.profiler.profile() as prof:
...     out = lucid.randn(128, 128) @ lucid.randn(128, 128)
>>> for row in prof.key_averages():
...     print(row)