fix(HuggingFace): drop unused TensorFlow import in sentiment analysis codegen

[kz930]

What changes were proposed in this PR?

HuggingFaceSentimentAnalysisOpDesc.generatePythonCode emitted an import for TFAutoModelForSequenceClassification that the generated body never uses — the model is loaded through the PyTorch AutoModelForSequenceClassification.

The symbol does not exist in the pinned transformers: the library dropped its TensorFlow backend in 5.0, and amber/operator-requirements.txt has been on 5.x since #5016 (5.5.0 today). So the import is not dead weight — it raises ImportError at module load and takes the whole generated script down before any inference runs, on every environment that installs the pinned requirements.

Fix — remove the unused import:

 from transformers import pipeline
 from transformers import AutoModelForSequenceClassification
-from transformers import TFAutoModelForSequenceClassification
 from transformers import AutoTokenizer, AutoConfig

Any related issues, documentation, discussions?

Closes #6794

How was this PR tested?

Manually, per @Yicong-Huang's request: the existing spec asserts on the emitted text and py_compile only parses it, so neither executes the template.

The template was dumped from generatePythonCode (columns text → pos/neu/neg) and run against the pytexera in amber/src/main/python, in a virtualenv with the pinned transformers 5.5.0 and numpy 2.1.0. main's stops at line 4 with ImportError: cannot import name 'TFAutoModelForSequenceClassification' from 'transformers', before the class body is defined; this PR's loads cardiffnlp/twitter-roberta-base-sentiment-latest and classifies all three rows, each row's three columns summing to 1:

text	pos	neu	neg
I love this new feature, it works beautifully!	0.9884	0.007	0.0046
The meeting is scheduled for 3pm in room 204.	0.0433	0.9438	0.0129
This is the worst release we have ever shipped.	0.0066	0.0423	0.951

Was this PR authored or co-authored using generative AI tooling?

Generated-by: Claude Code (Claude Opus 4.8)

[github-actions]

Automated Reviewer Suggestions

Based on the git blame history of the changed files, we recommend the following reviewers:

• Contributors with relevant context: @carloea2
  You can notify them by mentioning @carloea2 in a comment.

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 80.37%. Comparing base (5a3da70) to head (edf0679).
✅ All tests successful. No failed tests found.

Additional details and impacted files

@@             Coverage Diff              @@
##               main    #6809      +/-   ##
============================================
- Coverage     83.20%   80.37%   -2.83%     
- Complexity     4117     4691     +574     
============================================
  Files          1166     1222      +56     
  Lines         46456    50684    +4228     
  Branches       5177     5760     +583     
============================================
+ Hits          38654    40739    +2085     
- Misses         6096     8096    +2000     
- Partials       1706     1849     +143     

Flag	Coverage Δ		*Carryforward flag
access-control-service	71.10% <ø> (+1.10%)	⬆️
agent-service	83.65% <ø> (ø)		Carriedforward from ced81ee
amber	75.59% <ø> (-4.85%)	⬇️
computing-unit-managing-service	30.35% <ø> (-13.26%)	⬇️
config-service	64.02% <ø> (-1.96%)	⬇️
file-service	64.13% <ø> (-4.92%)	⬇️
frontend	83.55% <ø> (-0.29%)	⬇️	Carriedforward from ced81ee
notebook-migration-service	78.89% <ø> (ø)
pyamber	97.36% <ø> (-0.01%)	⬇️	Carriedforward from ced81ee
workflow-compiling-service	51.78% <ø> (+25.46%)	⬆️

*This pull request uses carry forward flags. Click here to find out more.

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[kz930]

@carloea2, @aglinxinyuan May any of you look at it please? Thank you.

[github-actions]

✅ No material benchmark regressions detected

🟢 0 better · 🔴 0 worse · ⚪ 15 noise (<±5%) · 0 without baseline

Compared against main 5a3da70 benchmarked on this same runner, so the delta is largely free of cross-runner hardware noise. The "7d avg" column still reflects the gh-pages dashboard. Treat <±5% as noise unless repeated.

Dashboard · Run

	config	throughput	MB/s	latency	max Δ latest / 7d
⚪	bs=10 sw=10 sl=64	450	0.274	20,708/30,940/30,940 us	⚪ within ±5% / 🔴 +99.1%
⚪	bs=100 sw=10 sl=64	972	0.593	100,306/125,128/125,128 us	⚪ within ±5% / 🔴 +14.7%
⚪	bs=1000 sw=10 sl=64	1,132	0.691	877,808/929,214/929,214 us	⚪ within ±5% / 🟢 -13.1%

Baseline details

Latest main 5a3da70 from same runner

config	metric	PR	latest main	7d avg	Δ latest	Δ 7d
bs=10 sw=10 sl=64	throughput	450 tuples/sec	472 tuples/sec	767.32 tuples/sec	-4.7%	-41.4%
bs=10 sw=10 sl=64	MB/s	0.274 MB/s	0.288 MB/s	0.468 MB/s	-4.9%	-41.5%
bs=10 sw=10 sl=64	p50	20,708 us	20,258 us	12,772 us	+2.2%	+62.1%
bs=10 sw=10 sl=64	p95	30,940 us	32,221 us	15,538 us	-4.0%	+99.1%
bs=10 sw=10 sl=64	p99	30,940 us	32,221 us	18,948 us	-4.0%	+63.3%
bs=100 sw=10 sl=64	throughput	972 tuples/sec	985 tuples/sec	972.51 tuples/sec	-1.3%	-0.1%
bs=100 sw=10 sl=64	MB/s	0.593 MB/s	0.601 MB/s	0.594 MB/s	-1.3%	-0.1%
bs=100 sw=10 sl=64	p50	100,306 us	98,842 us	103,020 us	+1.5%	-2.6%
bs=100 sw=10 sl=64	p95	125,128 us	126,123 us	109,070 us	-0.8%	+14.7%
bs=100 sw=10 sl=64	p99	125,128 us	126,123 us	118,964 us	-0.8%	+5.2%
bs=1000 sw=10 sl=64	throughput	1,132 tuples/sec	1,131 tuples/sec	1,005 tuples/sec	+0.1%	+12.6%
bs=1000 sw=10 sl=64	MB/s	0.691 MB/s	0.69 MB/s	0.613 MB/s	+0.1%	+12.7%
bs=1000 sw=10 sl=64	p50	877,808 us	880,420 us	1,002,400 us	-0.3%	-12.4%
bs=1000 sw=10 sl=64	p95	929,214 us	937,263 us	1,039,228 us	-0.9%	-10.6%
bs=1000 sw=10 sl=64	p99	929,214 us	937,263 us	1,069,081 us	-0.9%	-13.1%

Raw CSV

config_idx,batch_size,schema_width,string_len,num_batches,total_ms,total_tuples,total_bytes,tuples_per_sec,mb_per_sec,lat_p50_us,lat_p95_us,lat_p99_us
0,10,10,64,20,444.87,200,128000,450,0.274,20708.40,30940.25,30940.25
1,100,10,64,20,2057.43,2000,1280000,972,0.593,100305.90,125127.51,125127.51
2,1000,10,64,20,17673.46,20000,12800000,1132,0.691,877808.41,929214.26,929214.26

[Copilot]

Pull request overview

This PR fixes HuggingFace Sentiment Analysis Python code generation by removing an unused TensorFlow-only transformers import that can crash the generated script in PyTorch-only environments, aligning behavior with the operator’s actual PyTorch usage (Closes #6794).

Changes:

• Remove the unused TFAutoModelForSequenceClassification import from the generated Python code.
• Add a regression test ensuring the TF-only symbol is not emitted while the PyTorch AutoModelForSequenceClassification import remains present.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated no comments.

File	Description
common/workflow-operator/src/main/scala/org/apache/texera/amber/operator/huggingFace/HuggingFaceSentimentAnalysisOpDesc.scala	Drops the unused TensorFlow-only import from generatePythonCode() output.
common/workflow-operator/src/test/scala/org/apache/texera/amber/operator/huggingFace/HuggingFaceSentimentAnalysisOpDescSpec.scala	Adds regression coverage verifying the generated code no longer contains the TF-only import.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[aglinxinyuan]

Please fix the ci

[kz930]

CI green now — the title check was a transient failure, re-triggered by the new push. Also removed the test per @Yicong-Huang (4defe25). Thanks!

[Yicong-Huang]

CI green now — the title check was a transient failure, re-triggered by the new push. Also removed the test per @Yicong-Huang (4defe25). Thanks!

Please don't forget to update PR description ;)

[kz930]

Done, updated the description.

[Yicong-Huang]

Do we have a way to test and verify the generated python code can run as expected?

cc @carloea2 to see if you have context on this?

[aglinxinyuan]

Do we have a way to test and verify the generated python code can run as expected?

cc @carloea2 to see if you have context on this?

It's a separate topic, shouldn't block this PR.

I think currently all Python operators are untested by Python interpreter.

[Yicong-Huang]

If there are no programmatical tests, I will ask @kz930 to manually test and report the result in PR description. Thanks.

[carloea2]

@Yicong-Huang

I made a programmatic test to check all python operators generated code can py_compile. However it does not run them because of 2 reasons:

Increased run time.
How to feed meaningful smoke data.

Do you think is good idea to actually run them to fully test them?

[carloea2]

Also the test is supposed to check the operator correctly uses and don't forget to use Python template builder.

[Yicong-Huang]

@Yicong-Huang

I made a programmatic test to check all python operators generated code can py_compile. However it does not run them because of 2 reasons:

Increased run time. How to feed meaningful smoke data.

Do you think is good idea to actually run them to fully test them?

I think it will be great to have these added to CI! we can just ask each operator author/maintainer to provide a script to generate smoke data. Your test should provide an API to get data and get operator.

For this PR, if we have manual test, I am fine with it. We can push the programmatic test to future work. @carloea2 do you have an issue for it?

[carloea2]

I do not have an issue, that test was merged in the next PR: #4189

Do you want an issue for also running them?

[kz930]

Manually verified, numbers in the description: with the pinned transformers 5.5.0 the current template dies at ImportError: cannot import name 'TFAutoModelForSequenceClassification', while with this PR it classifies three rows correctly.

For a programmatic version — #7207 routes tests that actually run generated code to amber-integration, where torch and transformers are installed; the amber job installs no Python packages, which is why such a test has nowhere to live today.

[github-actions]

Backport auto-label report

This fix: PR was checked against each actively-supported release branch. release/* labels drive the post-merge backport, so add or remove one to change where this fix lands.

Release branch	Analysis
✅ release/v1.2	Change detected on this branch — label added; this fix is queued to backport here. Requested review from @xuang7.

Auto-label run.

[github-actions]

Backport PR opened: draft #7315 (#7315) to release/v1.2, assigned to @kz930 — needs manual work because the backported tree failed its pre-merge build.