[
https://issues.apache.org/jira/browse/BEAM-13983?focusedWorklogId=763906&page=com.atlassian.jira.plugin.system.issuetabpanels:worklog-tabpanel#worklog-763906
]
ASF GitHub Bot logged work on BEAM-13983:
-----------------------------------------
Author: ASF GitHub Bot
Created on: 28/Apr/22 21:37
Start Date: 28/Apr/22 21:37
Worklog Time Spent: 10m
Work Description: ryanthompson591 commented on PR #17496:
URL: https://github.com/apache/beam/pull/17496#issuecomment-1112676641
@TheNeuralBit Originally, I felt the same. sklearn and pytorch, let me know
if I'm wrong. The problem I was seeing was name collision.
For example if someone had the following imports.
import sklearn
from apache_beam.ml.inference import sklearn
there would be a name collision. Problem is that python does raise an
exception here, but simply has the last import stand. So this will cause our
users debugging head aches.
Instead they would need to do something like.
import sklearn
import apache_beam.ml.inference.sklearn as apache_sklearn (or
sklearn_inference or whatever)
Our feeling is that users will often also import sklearn to use some of the
libraries there. So in order to avoid error, we plan to use
framework_inference as the name. I liked loader and andy liked inference. So
we flipped a coin.
Issue Time Tracking
-------------------
Worklog Id: (was: 763906)
Time Spent: 8.5h (was: 8h 20m)
> Implement RunInference for Scikit-learn
> ---------------------------------------
>
> Key: BEAM-13983
> URL: https://issues.apache.org/jira/browse/BEAM-13983
> Project: Beam
> Issue Type: Sub-task
> Components: sdk-py-core
> Reporter: Andy Ye
> Assignee: Ryan Thompson
> Priority: P2
> Labels: run-inference
> Time Spent: 8.5h
> Remaining Estimate: 0h
>
> Implement RunInference for Scikit-learn as described in the design doc
> [https://s.apache.org/inference-sklearn-pytorch]
> There will be a sklearn_impl.py file that contains SklearnModelLoader and
> SlkearnInferenceRunner classes.
--
This message was sent by Atlassian Jira
(v8.20.7#820007)
