[Question] Custom Architecture

[hg0428]

❓ General Questions

Hi, I have a custom pretrained LLM with a custom non-standard architecture implemented in Pytorch.
Gemini Deep Research told me that MLC LLM can compile any architecture to be able to run on MLC LLM. Is that true? If so, can you point me to the documentation on how to do it?
If not, would a solution such as Apache TVM be good for cross-platform inference (including mobile)?