What is the maximum model size supported by core ml ?

## ❓Question

I have been trying to convert flan-t5 transformer models, which are more than 4 gb in size. Coreml conversion crashes in this case. I have a 64 GB M1 Max system.

Is the an observed behavior ? Or is it just me ?

No source files, tests, conversion command, version details, or crash output are provided. Start by reproducing the Flan-T5 conversion on the reported M1 Max environment and inspect the failure; done means documenting whether conversion above 4 GB is expected and identifying any supported model-size limit.

Written by the indexing model from the issue text.