[Python] Support loading of TF models with saved weights #25496

riteshghorse · 2023-02-15T20:51:52Z

Adds support for loading tensorflow model from saved weights given a function to create the model.
Follow up to #25368, Closes #25366

Thank you for your contribution! Follow this checklist to help us incorporate your contribution quickly and easily:

Mention the appropriate issue in your description (for example: addresses #123), if applicable. This will automatically add a link to the pull request in the issue. If you would like the issue to automatically close on merging the pull request, comment fixes #<ISSUE NUMBER> instead.
Update CHANGES.md with noteworthy changes.
If this contribution is large, please file an Apache Individual Contributor License Agreement.

See the Contributor Guide for more tips on how to make review process smoother.

To check the build health, please visit https://github.com/apache/beam/blob/master/.test-infra/BUILD_STATUS.md

GitHub Actions Tests Status (on master branch)

See CI.md for more information about GitHub Actions CI.

codecov · 2023-02-15T21:35:14Z

Codecov Report

Merging #25496 (deaa4df) into master (921bc7b) will increase coverage by 0.60%.
The diff coverage is 0.00%.

@@            Coverage Diff             @@
##           master   #25496      +/-   ##
==========================================
+ Coverage   72.20%   72.81%   +0.60%     
==========================================
  Files         772      751      -21     
  Lines      102264    99590    -2674     
==========================================
- Hits        73838    72514    -1324     
+ Misses      26992    25715    -1277     
+ Partials     1434     1361      -73

Flag	Coverage Δ
python	`81.98% <0.00%> (-0.03%)`	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

Impacted Files	Coverage Δ
...xamples/inference/tensorflow_mnist_with_weights.py	`0.00% <0.00%> (ø)`
...n/apache_beam/ml/inference/tensorflow_inference.py	`0.00% <0.00%> (ø)`
...ks/go/pkg/beam/runners/dataflow/dataflowlib/job.go	`24.59% <0.00%> (-4.71%)`	⬇️
sdks/go/pkg/beam/io/filesystem/gcs/gcs.go	`7.29% <0.00%> (-4.17%)`	⬇️
sdks/go/pkg/beam/core/runtime/exec/sdf_invokers.go	`72.14% <0.00%> (-4.17%)`	⬇️
sdks/python/apache_beam/utils/interactive_utils.py	`95.12% <0.00%> (-2.44%)`	⬇️
sdks/go/pkg/beam/core/funcx/output.go	`85.71% <0.00%> (-1.25%)`	⬇️
sdks/go/pkg/beam/io/parquetio/parquetio.go	`59.55% <0.00%> (-0.89%)`	⬇️
.../apache_beam/runners/direct/transform_evaluator.py	`89.57% <0.00%> (-0.76%)`	⬇️
sdks/go/pkg/beam/core/graph/fn.go	`84.40% <0.00%> (-0.71%)`	⬇️
... and 35 more

📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more

riteshghorse · 2023-02-16T17:29:47Z

Python_PVR_Flink failed because of licensing issue. Bruno has a PR at #25515

riteshghorse · 2023-02-16T18:27:10Z

R: @damccorm @jrmccluskey

riteshghorse · 2023-02-16T19:04:13Z

Run Python_PVR_Flink PreCommit

sdks/python/apache_beam/ml/inference/tensorflow_inference.py

damccorm · 2023-02-16T20:11:55Z

Run Python 3.8 PostCommit

riteshghorse · 2023-02-17T14:02:42Z

Run Python 3.8 PostCommit

damccorm · 2023-02-17T14:34:26Z

I think you need to pull in master to avoid postcommit failures (fixed by #25446)

…model-wt

riteshghorse · 2023-02-17T14:42:01Z

Run Python 3.8 PostCommit

riteshghorse · 2023-02-17T19:37:06Z

Run Python_Dataframes PreCommit

riteshghorse · 2023-02-17T20:57:45Z

Run Whitespace PreCommit

sdks/python/tox.ini

riteshghorse · 2023-02-17T21:22:15Z

Run Python 3.8 PostCommit

github-actions · 2023-02-17T22:34:54Z

Assigning reviewers. If you would like to opt out of this review, comment assign to next reviewer:

R: @tvalentyn for label python.

Available commands:

stop reviewer notifications - opt out of the automated review tooling
remind me after tests pass - tag the comment author after tests pass
waiting on author - shift the attention set back to the author (any comment or push by the author will return the attention set to the reviewers)

The PR bot will only process comments in the main thread (not review comments).

tvalentyn · 2023-02-18T01:52:20Z

sdks/python/setup.py

@@ -286,6 +286,7 @@ def get_portability_package_data():
            'pytest-xdist>=2.5.0,<3',
            'pytest-timeout>=2.1.0,<3',
            'scikit-learn>=0.20.0',
+            'tensorflow>=1.0.0',


(drive-by comment): let's not depend on a large dependency like this by default. you can see how we deal with sklearn/pytorch in similar circumstances.

+1 to @tvalentyn.

Commented at #25547 (comment).

We can follow the approach of sklearn/torch/tensorRT gradle tests and not add very large ML dependencies to the setup.py

Ah, I see - so this was running on 3.7 and 3.9, just not 3.8 because of the difference in dependencies here -

beam/build.gradle.kts

Line 457 in 921bc7b

dependsOn(":sdks:python:test-suites:direct:py37:inferencePostCommitIT")

Thanks for clarifying. I'm now +1 on removing this and manually triggering Python 3.7 postcommits, sorry for my confusion @riteshghorse

A couple of follow up questions for @AnandInguva and @tvalentyn :

Is there any reason we only run the dataflow versions on 3.8? Since TensorRT is on Dataflow, we now need to run both the 3.7 and 3.8 postcommits to fully exercise our inference code base

Is there any reason we're not running most of our inference postcommits on dataflow (as well as the direct runner)? That's probably where most of our usage is today

TensoRT tests run on Python 3.8 because of the container image provided by TensorRT folks contains 3.8. We need to have an automated process of building image from Dockerfile for Python 3.7, 3.9, 3.10 and then use that image for the tests in Dataflow.

Initially, these tests are very light weight IT tests, so we just run on the DirectRunner. apart from that, I don't see any reason why we didn't run on Dataflow. Maybe adding these postcommit tests to Dataflow suite would increase the total time of Dataflow suite tests in the PostCommit suite.

riteshghorse · 2023-02-21T16:06:18Z

Run Python 3.7 PostCommit

riteshghorse · 2023-02-21T16:44:45Z

Run Python Unit Tests

riteshghorse · 2023-02-21T16:52:28Z

Run Python 3.9 PostCommit

damccorm · 2023-02-21T18:28:34Z

3.7 has passing tensorflow tests. I'll wait until all active suites complete and merge

damccorm · 2023-02-21T18:28:59Z

(or you can merge)

damccorm · 2023-02-21T21:47:45Z

Run Python_Coverage PreCommit

damccorm · 2023-02-21T21:48:01Z

(all other checks have passed, they're just not statusing)

riteshghorse · 2023-02-22T00:12:34Z

All checks passed. Merging!

* load model with weight * example * update test * update test * make create model fn optional * change tf to tensorflow * add readme and change urls * fix whitespace * add doc and changes.md * add tensorflow dependency * remove tf dependency

riteshghorse added 4 commits February 15, 2023 15:49

load model with weight

237f33a

example

ff552be

update test

4961209

update test

69147b4

riteshghorse added 2 commits February 15, 2023 16:51

make create model fn optional

894d66b

change tf to tensorflow

6240c40

riteshghorse marked this pull request as ready for review February 16, 2023 03:14

riteshghorse marked this pull request as draft February 16, 2023 03:15

add readme and change urls

13a2023

riteshghorse marked this pull request as ready for review February 16, 2023 15:27

riteshghorse changed the title ~~[Python] Support loading models with saved weights~~ [Python] Support loading of TF models with saved weights Feb 16, 2023

fix whitespace

e9b8301

damccorm reviewed Feb 16, 2023

View reviewed changes

sdks/python/apache_beam/ml/inference/tensorflow_inference.py Show resolved Hide resolved

add doc and changes.md

1ff33ea

damccorm mentioned this pull request Feb 17, 2023

[Python] Added Tensorflow Model Handler #25368

Merged

3 tasks

Merge branch 'master' of https://github.com/apache/beam into tf-load-…

313c8b0

…model-wt

damccorm reviewed Feb 17, 2023

View reviewed changes

sdks/python/tox.ini Show resolved Hide resolved

add tensorflow dependency

5c8f6f5

github-actions bot added examples python labels Feb 17, 2023

damccorm added a commit that referenced this pull request Feb 17, 2023

Port change from #25496

fccffaf

damccorm mentioned this pull request Feb 17, 2023

Add dependencies needed for some ml integration tests #25548

Closed

3 tasks

github-actions bot added the Next Action: Reviewers label Feb 17, 2023

tvalentyn reviewed Feb 18, 2023

View reviewed changes

riteshghorse added 2 commits February 21, 2023 10:57

merge master

6bd5d47

remove tf dependency

deaa4df

damccorm approved these changes Feb 21, 2023

View reviewed changes

riteshghorse merged commit 33750c1 into apache:master Feb 22, 2023

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

[Python] Support loading of TF models with saved weights #25496

[Python] Support loading of TF models with saved weights #25496

riteshghorse commented Feb 15, 2023 •

edited

Loading

codecov bot commented Feb 15, 2023 •

edited

Loading

riteshghorse commented Feb 16, 2023

riteshghorse commented Feb 16, 2023

riteshghorse commented Feb 16, 2023

damccorm commented Feb 16, 2023

riteshghorse commented Feb 17, 2023

damccorm commented Feb 17, 2023

riteshghorse commented Feb 17, 2023

riteshghorse commented Feb 17, 2023

riteshghorse commented Feb 17, 2023

riteshghorse commented Feb 17, 2023

github-actions bot commented Feb 17, 2023

tvalentyn Feb 18, 2023 •

edited

Loading

AnandInguva Feb 18, 2023

AnandInguva Feb 18, 2023

damccorm Feb 21, 2023

AnandInguva Feb 21, 2023

riteshghorse commented Feb 21, 2023

riteshghorse commented Feb 21, 2023

riteshghorse commented Feb 21, 2023

damccorm commented Feb 21, 2023

damccorm commented Feb 21, 2023

damccorm commented Feb 21, 2023

damccorm commented Feb 21, 2023

riteshghorse commented Feb 22, 2023

[Python] Support loading of TF models with saved weights #25496

[Python] Support loading of TF models with saved weights #25496

Conversation

riteshghorse commented Feb 15, 2023 • edited Loading

GitHub Actions Tests Status (on master branch)

codecov bot commented Feb 15, 2023 • edited Loading

Codecov Report

riteshghorse commented Feb 16, 2023

riteshghorse commented Feb 16, 2023

riteshghorse commented Feb 16, 2023

damccorm commented Feb 16, 2023

riteshghorse commented Feb 17, 2023

damccorm commented Feb 17, 2023

riteshghorse commented Feb 17, 2023

riteshghorse commented Feb 17, 2023

riteshghorse commented Feb 17, 2023

riteshghorse commented Feb 17, 2023

github-actions bot commented Feb 17, 2023

tvalentyn Feb 18, 2023 • edited Loading

Choose a reason for hiding this comment

AnandInguva Feb 18, 2023

Choose a reason for hiding this comment

AnandInguva Feb 18, 2023

Choose a reason for hiding this comment

damccorm Feb 21, 2023

Choose a reason for hiding this comment

AnandInguva Feb 21, 2023

Choose a reason for hiding this comment

riteshghorse commented Feb 21, 2023

riteshghorse commented Feb 21, 2023

riteshghorse commented Feb 21, 2023

damccorm commented Feb 21, 2023

damccorm commented Feb 21, 2023

damccorm commented Feb 21, 2023

damccorm commented Feb 21, 2023

riteshghorse commented Feb 22, 2023

riteshghorse commented Feb 15, 2023 •

edited

Loading

codecov bot commented Feb 15, 2023 •

edited

Loading

tvalentyn Feb 18, 2023 •

edited

Loading