[Feature]: Add TRTLLM as an attention backend for Auto Deploy

[bmarimuthu-nv]

🚀 The feature, motivation and pitch

TRTLLM attention backend gives better perf for some models. Until flashInfer catches up, it'll be good to use TRTLLM backend to have on-par perf. So we need to add this backend option.

Alternatives

No response

Additional context

No response

Before submitting a new issue...

• Make sure you already searched for relevant issues, and checked the documentation and examples for answers to frequently asked questions.

[coderabbitai]

📝 CodeRabbit Plan Mode

Generate an implementation plan and prompts that you can use with your favorite coding agent.

• Create Plan

Examples

• Example 1
• Example 2

---

🔗 Similar Issues

Related Issues

• [AutoDeploy]: Fix the size of the memory used during forward when resizing the kv-cache #8626
• [Feature]: AutoDeploy: multi-stream MoE support #9271
• [AutoDeploy] Basic support with model build + cached attention + multi-gpu support #7308
• [Feature]: AutoDeploy: Set up the Accuracy testing for Nemotron MOE models. #9316
• [Feature]: AutoDeploy: optimize fp8+TP8 perf for nano3 #9273

👤 Suggested Assignees

• lucaslie
• nzmora-nvidia
• nvchenghaoz

---

🧪 Issue enrichment is currently in open beta.

You can configure auto-planning by selecting labels in the issue_enrichment configuration.

To disable automatic issue enrichment, add the following to your .coderabbit.yaml:

issue_enrichment:
  auto_enrich:
    enabled: false

💬 Have feedback or questions? Drop into our discord or schedule a call!