[None][docs] fix incorrect auto sampler behavior description for beam search

[fuergaosi233]

Summary

• docs/source/features/sampling.md incorrectly stated that the auto sampler mode uses TRTLLMSampler when beam search is enabled
• In practice, auto always selects TorchSampler for all requests, including beam search
• This contradicts both the field description in llm_args.py:3810 ("Defaults to auto, which will use TorchSampler") and the actual sampler instantiation logic in _torch/pyexecutor/_util.py:1821

Root Cause

The auto mode was historically documented as routing to TRTLLMSampler for beam search, but this behavior does not exist in the code. TRTLLMSampler is only used when explicitly set via sampler_type="TRTLLMSampler", which is deprecated and scheduled for removal in release 1.4.

Code evidence:

• tensorrt_llm/_torch/pyexecutor/_util.py:1821: Only checks if llm_args.sampler_type == SamplerType.TRTLLMSampler — no auto-to-TRTLLM path for beam search
• tensorrt_llm/llmapi/llm_args.py:3810: Field description says "Defaults to auto, which will use TorchSampler"
• tensorrt_llm/_torch/auto_deploy/shim/ad_executor.py:843-844: if sampler_type == SamplerType.auto: sampler_type = SamplerType.TorchSampler

Test plan

• Verify the updated description is consistent with llm_args.py field description
• Verify no code path exists where auto + beam search selects TRTLLMSampler

Summary by CodeRabbit

• Documentation
  • Updated sampling backend documentation to clarify that the auto mode defaults to using Torch Sampler for all requests, providing consistent behavior across sampling configurations.

[coderabbitai]

📝 Walkthrough

Walkthrough

This PR updates the sampling backend documentation to clarify that the default auto selection uses Torch Sampler for all requests, removing the prior conditional behavior that selected TRTLLM Sampler for Beam Search operations.

Changes

Sampling Backend Documentation

Layer / File(s)	Summary
Sampling backend default behavior documentation docs/source/features/sampling.md	The default auto sampling backend documentation now explicitly states that Torch Sampler is used for all requests, replacing the previous conditional description that selected TRTLLM Sampler when using Beam Search.

🎯 1 (Trivial) | ⏱️ ~2 minutes

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title clearly describes the main change: fixing incorrect documentation about auto sampler behavior for beam search.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Description check	✅ Passed	The PR description comprehensively explains the issue, root cause, and provides specific code evidence with line references.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests

---

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[Funatiq]

Good catch. Somehow we missed that.

[Funatiq]

/bot skip --comment "Changed only docs"

[tensorrt-cicd]

PR_Github #50727 [ skip ] triggered by Bot. Commit: 8612c60 Link to invocation

[tensorrt-cicd]

PR_Github #50727 [ skip ] completed with state SUCCESS. Commit: 8612c60
Skipping testing for commit 8612c60

Link to invocation