[BugFix][DataProcessor] Force top_k=1 for greedy decoding when temperature=0

[gongweibao]

Motivation

When temperature=0 (greedy decoding), the current code only sets temperature to a small epsilon (1e-06). However, this is insufficient — the sampling kernel may still pick non-top-1 tokens due to floating-point noise in the softmax output. This causes non-deterministic outputs for greedy decoding requests.

This PR explicitly sets top_k=1 alongside the temperature epsilon to guarantee argmax behavior.

Modifications

• fastdeploy/engine/engine.py: Add sampling_params.top_k = 1 when temperature ≈ 0.
• fastdeploy/input/text_processor.py: Add top_k=1 in both process_request and _process_request paths.
• fastdeploy/input/v1/text_processor.py: Same as above for v1 processor.
• fastdeploy/input/ernie4_5_processor.py: Same as above for Ernie4.5 processor.
• tests/input/test_text_processor.py: Add top_k=1 assertion in existing greedy decoding tests.

Usage or Command

No new usage. Existing requests with temperature=0 will now automatically use top_k=1 for true greedy decoding.

Accuracy Tests

This is a sampling parameter fix, not a kernel/model change. The fix ensures greedy decoding always selects the top-1 token, which is the expected behavior when temperature=0.

Checklist

• Add at least a tag in the PR title: [BugFix], [Engine], [DataProcessor]
• Format your code, run pre-commit before commit.
• Add unit tests: tests/input/test_text_processor.py updated.
• Provide accuracy results: N/A (sampling parameter fix, not model output change).
• If the current PR is submitting to the release branch, make sure the PR has been submitted to the develop branch. (N/A — targeting develop)

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.

---

gongweibao seems not to be a GitHub user. You need a GitHub account to be able to sign the CLA. If you have already a GitHub account, please add the email address used for this commit to your account.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}

[paddle-bot]

Thanks for your contribution!

[codecov-commenter]

Codecov Report

❌ Patch coverage is 82.60870% with 4 lines in your changes missing coverage. Please review.
⚠️ Please upload report for BASE (develop@28f7727). Learn more about missing BASE report.

Files with missing lines	Patch %	Lines
...executor/layers/sample/ops/top_k_top_p_sampling.py	80.00%	1 Missing and 2 partials ⚠️
fastdeploy/input/v1/text_processor.py	50.00%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             develop    #6748   +/-   ##
==========================================
  Coverage           ?   71.94%           
==========================================
  Files              ?      392           
  Lines              ?    53903           
  Branches           ?     8476           
==========================================
  Hits               ?    38779           
  Misses             ?    12348           
  Partials           ?     2776           

Flag	Coverage Δ
GPU	71.94% <82.60%> (?)

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

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

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.