Add triton runtime

[fyuan1316]

Summary by CodeRabbit

• Documentation
  • Added Triton Inference Server as a new runtime example with full configuration and step-by-step usage for NVIDIA GPU deployments.
  • Updated the runtime comparisons table to include Triton's target hardware, supported frameworks, and configuration notes.

_{✏️ Tip: You can customize this high-level summary in your review settings.}

[coderabbitai]

Walkthrough

Adds a Triton Inference Server example to the custom inference runtime docs, including a full ClusterServingRuntime YAML configured for NVIDIA GPUs, startup commands/env vars, resource settings, startupProbe, supportedModelFormats, usage steps, and an update to the runtime comparison table.

Changes

Cohort / File(s)	Summary
Documentation: Triton Inference Server Runtime docs/en/model_inference/inference_service/how_to/custom_inference_runtime.mdx	Adds a new "Specific Runtime Examples" entry for Triton with a full ClusterServingRuntime YAML (GPU accelerator metadata, container args/env, resources, startupProbe, supportedModelFormats), usage/preparation steps, and a new row in the runtime comparison table.

Sequence Diagram(s)

(omitted)

Estimated code review effort

🎯 1 (Trivial) | ⏱️ ~5 minutes

Possibly related PRs

• Update custom infer runtime docs #35: Modifies the same custom_inference_runtime.mdx to add/expand custom runtime examples (related additions for other runtimes).
• Add ascend 310p runtime #78: Updates custom_inference_runtime.mdx with additional specific runtime examples and comparison table edits.
• AI-22071 Add MLServer runtime #14: Adds other custom runtime YAML examples to the same documentation page.

Suggested reviewers

• typhoonzero

Poem

🐰 I hopped through YAML, neat and spry,
Triton lights the GPU sky,
Containers, probes, and models align,
Docs now hum with runtime shine,
🥕🚀

🚥 Pre-merge checks | ✅ 3

✅ Passed checks (3 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title 'Add triton runtime' directly describes the main change: adding documentation for Triton Inference Server runtime with YAML configuration and usage steps.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.

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

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

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

[coderabbitai]

Actionable comments posted: 1

🤖 Fix all issues with AI agents

In
`@docs/en/model_inference/inference_service/how_to/custom_inference_runtime.mdx`:
- Around line 329-333: Add a Kubernetes startupProbe entry to the Triton runtime
YAML so the pod is not considered ready until the model finishes loading;
specifically, insert a startupProbe block (mirroring the pattern used in other
runtimes) immediately before the supportedModelFormats section in the Triton
runtime example (near the runAsUser key and before supportedModelFormats: -
name: triton) to probe the model server endpoint until it is healthy.

🧹 Nitpick comments (2)

docs/en/model_inference/inference_service/how_to/custom_inference_runtime.mdx (2)

308-312: Unused environment variable MODEL_REPO.

The MODEL_REPO environment variable is defined on lines 311-312 but is not used anywhere in the container command (lines 302-307). Either remove it or use it in the command if it's intended for some purpose.

🔧 Suggested fix: Remove unused environment variable

      env:
        - name: OMP_NUM_THREADS
          value: "1"
-        - name: MODEL_REPO
-          value: '{{ index .Annotations "aml-model-repo" }}'
      image: 152-231-registry.alauda.cn:60070/mlops/tritonserver:25.02-py3

---

313-313: Internal registry image may not be accessible to users.

The image 152-231-registry.alauda.cn:60070/mlops/tritonserver:25.02-py3 appears to reference an internal registry. Consider adding a comment similar to other examples, or use the official NVIDIA NGC image reference (e.g., nvcr.io/nvidia/tritonserver:25.02-py3) for better accessibility.

🔧 Suggested fix: Use official NVIDIA image

-      image: 152-231-registry.alauda.cn:60070/mlops/tritonserver:25.02-py3
+      image: nvcr.io/nvidia/tritonserver:25.02-py3  # Replace with your actual image if needed

[cloudflare-workers-and-pages]

Deploying alauda-ai with    Cloudflare Pages

Latest commit:	217ba40
Status:	✅  Deploy successful!
Preview URL:	https://49a3ab18.alauda-ai.pages.dev
Branch Preview URL:	https://add-triton-rt.alauda-ai.pages.dev

View logs

[liuwei-2622]

/test-pass