feat: Expose BF16 precision in TensorRT

[dwyatte]

What does the PR do?

BF16 was added to the ONNX runtime TensorRT EP in microsoft/onnxruntime#24743, this PR should expose it to Triton's ONNX backend

Checklist

• I have read the Contribution guidelines and signed the Contributor License
  Agreement
• PR title reflects the change and is of format <commit_type>: <Title>
• Changes are described in the pull request.
• Related issues are referenced.
• Populated github labels field
• [n/a] Added test plan and verified test passes.
• Verified that the PR passes existing CI.
• I ran pre-commit locally (pre-commit install, pre-commit run --all)
• Verified copyright is correct on all changed files.
• [n/a] Added succinct git squash message before merging ref.
• All template sections are filled out.
• [n/a] Optional: Additional screenshots for behavior/output changes with before/after.

Commit Type:

Check the conventional commit type
box here and add the label to the github PR.

• build
• ci
• docs
• feat
• fix
• perf
• refactor
• revert
• style
• test

Related PRs:

n/a

Where should the reviewer start?

src/onnxruntime.cc

Test plan:

n/a

Caveats:

n/a

Background

See microsoft/onnxruntime#24743 for more info

Related Issues: (use one of the action keywords Closes / Fixes / Resolves / Relates to)

• closes GitHub issue: BF16 support for integrated TensorRT precision mode server#5959

[whoisj]

Adding @yinggeh to review this. I'm not as familiar w/ ONNX->TRT as I should be.

[yinggeh]

Adding @yinggeh to review this. I'm not as familiar w/ ONNX->TRT as I should be.

Not familiar either but looks like a small change. I will take a look today.

[whoisj]

@dwyatte, please make sure you've completed the contribution requirements: https://github.com/triton-inference-server/server?tab=readme-ov-file#contributing. Thank you.

[dwyatte]

@dwyatte, please make sure you've completed the contribution requirements: https://github.com/triton-inference-server/server?tab=readme-ov-file#contributing. Thank you.

@whoisj Block (my corporate entity) has previously completed the CLA here, but let me know if I need to personally submit something too

[yinggeh]

LGTM if the change is indeed this simple.

[yinggeh]

Please update PR title and description using the template https://github.com/triton-inference-server/server/blob/main/.github/PULL_REQUEST_TEMPLATE/pull_request_template_external_contrib.md. Fill n/a for any field doesn't apply

[dwyatte]

Please update PR title and description using the template https://github.com/triton-inference-server/server/blob/main/.github/PULL_REQUEST_TEMPLATE/pull_request_template_external_contrib.md. Fill n/a for any field doesn't apply

@yinggeh Done!

[yinggeh]

@whoisj Is our client eligible for contributing now?

[yinggeh]

@whoisj Can I merge?

[whoisj]

@yinggeh as far as I know, yes.