Add --min-deployment-target Argument to Set Deployment Target

[swiftlysingh]

Hi everyone,

I’m working on converting the Stable Diffusion XL model for iOS 18 and encountered an issue during post-training quantization. The error message suggested that iOS 16 doesn't support the optimizations I'm applying:

AssertionError: The IOS16 only supports per-tensor LUT, but got more than one LUT on 0th axis. LUT shape: (80, 1, 1, 1, 16, 1)

This seems to be due to the minimum deployment target being set to iOS 16 by default. Updating the target to iOS 18 should resolve the issue.

For more context, see this related Apple Developer forum thread.

Proposed Solution:

To make this more flexible, I propose adding a --min-deployment-target argument to the conversion script, which would allow users to set their desired deployment target. This could be set to the current default macOS15.

Open to Suggestions:

I’m open to feedback and would appreciate any input on the best way to implement this feature. If there’s a preferred approach for handling deployment target settings, I’d be happy to discuss further or explore alternatives. Let me know if you'd like me to submit a PR, or if there are any changes you’d suggest to make this more robust.

Best regards,
Pushpinder Pal Singh