Run out of disk spaces when converting models.

I am using coremltools to develop models for Mac. I have to convert many models and then use model.predict to get the evaluation time. But I found that I'm running out of space on my Mac's disk after some tests. Sometimes rebooting the mac could solve. But I still do not know what I should do to avoid it. Please help!

The report names repeated coremltools model conversion and model.predict on a Mac but no source file or test. Reproduce the repeated conversion and prediction workflow while monitoring disk usage, then identify the responsible operation and add a regression test or documented mitigation showing that repeated runs no longer consume unreleased disk space.

Written by the indexing model from the issue text.