Question: Interest in HeartMuLa 3B music-generation support?

Topic

Architecture or contribution guidance

Question

Hi @yifeif-nv, would HeartMuLa 3B music-generation support be a welcome contribution to TensorRT Model Connect? After working on SmolLM3 in #1158, I would be interested in exploring a family-local integration of HeartMuLa and HeartCodec, starting with a feasibility check before committing to the full implementation.

The public models generate music from lyrics and style tags. I have access to H100/H200 GPUs for validation. Is this a direction you would be interested in, and is anyone already working on it?

Model

• HeartMuLa-oss-3B-happy-new-year
• HeartCodec-oss-20260123
• Official implementation

What have you tried?

I checked the current family list and searched existing issues and PRs; I did not find HeartMuLa support. No TensorRT build or performance qualification has been attempted yet. Exact checkpoint revisions would be pinned for the feasibility work.

Submission checks

• I searched the documentation and open and closed issues first.
• This question contains no secrets, private URLs, or restricted artifacts.

Review the current family list and the prior SmolLM3 work in #1158, then compare HeartMuLa and HeartCodec with the linked official implementation. Confirm the exact checkpoint revisions and assess feasibility on H100/H200 GPUs; done means maintainers clarify whether the contribution is welcome and whether anyone is already working on it.