[Feature]: Support Minimax model in AutoDeploy

[suyoggupta]

🚀 The feature, motivation and pitch

https://huggingface.co/MiniMaxAI/MiniMax-M2/tree/main

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

• [Feature]: AutoDeploy: optimize fp8+TP8 perf for nano3 #9273
• [Feature]: AutoDeploy: check TRTLLMSampler on host-bound test cases (B200, H200, tp8, fp8, ...) #9846
• [Feature]: AutoDeploy: multi-stream MoE support #9271
• [Feature]: AutoDeploy: Set up the Accuracy testing for Nemotron MOE models. #9316
• [AutoDeploy]: Fix the size of the memory used during forward when resizing the kv-cache #8626

👤 Suggested Assignees

• lucaslie
• nvchenghaoz
• nzmora-nvidia

---

🧪 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!

[bmarimuthu-nv]

Fixes:

• issue with weights dtype in the traced graph (fp32) however checkpoint weights are in bf16. config.json didn't have explicit dtype.
• QKV weight split detection detection had a bug
• QK norm weights not getting same sharding applied