streaming.base.shared#
Objects that live in shared memory.
For when using threading or multiprocessing from the python standard library won’t do, because we are coordinating separately instantiated pytorch worker processes.
Functions
Register or lookup our shared memory prefix. |
Classes
A numpy array of predetermined shape and dtype that lives in shared memory. |
|
A barrier that works inter-process using a filelock and shared memory. |
|
Improved quiet implementation of shared memory. |
|
A numpy scalar of predetermined dtype that lives in shared memory. |