You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Although it is possible to support parallelism of triage components by using specific triage components instead of the whole pipeline, it should also support some simple version of this out of the box.
An injected executor could be a good way to do this (ie SimpleExecutor, MultiprocessingExecutor).
A better first pass, now that pipeline.run() is more compact, would be to subclass Pipeline and only reimplement .run().
The text was updated successfully, but these errors were encountered:
- Create abstract class PipelineBase, which sets up components and requires subclasses to implement run
- Move current Pipeline to SerialPipeline, subclass of PipelineBase
- Create LocalParallelPipeline subclass of PipelineBase to spawn workers using concurrent.futures.ProcessPoolExecutor. Currently just parallelizing the testing phase
- Make poolclass configurable, as NullPool was giving some problems with multiple processes and QueuePool seems to be more robust
- Modify MettaCSVMatrixStorage to only load data upon use, improving serializability
- Fix logging argument error in IndividualImportance (it doesn't raise exception, but does show up as an error in the log)
- Fix replace logic [Resolves#28]
Although it is possible to support parallelism of triage components by using specific triage components instead of the whole pipeline, it should also support some simple version of this out of the box.
An injected executor could be a good way to do this (ie SimpleExecutor, MultiprocessingExecutor).
A better first pass, now that pipeline.run() is more compact, would be to subclass Pipeline and only reimplement .run().
The text was updated successfully, but these errors were encountered: