Support quatization of loadConstantND

[Ghost-LZW]

No description provided.

[TobyRoseman]

@Ghost-LZW - Please include a unit test for this fix, i.e. a unit test which fails without your fix but passes with your fix. Thank you.

[Ghost-LZW]

@Ghost-LZW - Please include a unit test for this fix, i.e. a unit test which fails without your fix but passes with your fix. Thank you.

The corresponding unit test has been added

[TobyRoseman]

Thanks for adding the unit test. This change looks good to me.

I have kicked off a CI run for this change:
https://gitlab.com/coremltools1/coremltools/-/pipelines/956599663

If that passes, I plan to merge this pull request.

[Ghost-LZW]

Thanks for adding the unit test. This change looks good to me.

I have kicked off a CI run for this change: https://gitlab.com/coremltools1/coremltools/-/pipelines/956599663

If that passes, I plan to merge this pull request.

failed case is coremltools.test.neural_network.test_tf_numeric.StressTest, it seems like nothing about quantization_utils. When I use the main branch to run UT locally, also get failed (with TensorFlow 2.11.0 python10 on m2 pro). How can I fix it or some issue in ci?

[TobyRoseman]

I restarted the failed job and it failed again. However this test is passing in main. So I think this must be related to your change.

A PSNR 8 is quite poor. I suggest you dig into what this unit test is doing.

[TobyRoseman]

Changes look good. Thanks.

Updated CI: https://gitlab.com/coremltools1/coremltools/-/pipelines/988268661

If CI passes, I will merge.

[TobyRoseman]

@Ghost-LZW - I'm still seeing the same PSNR Depthwise Convolution failure on the updated CI run.

[Ghost-LZW]

@TobyRoseman sorry for late reply. Busy for other things recently, I'll dig into this unit test until I can give it some time.

[TobyRoseman]

Thanks for adding unit tests.

I've kicked off a CI for this change:
https://gitlab.com/coremltools1/coremltools/-/pipelines/1104510088

[Ghost-LZW]

Over the last weekend, I attempted to rebase to origin/main in order to investigate the reason behind the failure of PSNR Depthwise Convolution. However, I was unable to successfully build the CI environment, so I haven't been able to do anything yet. Later on, the CI process passed, which seemed a bit strange to me.

[TobyRoseman]

The most recent CI run is good. Don't worry about the previous CI. It's likely unrelated to your change.

If you can remove that print statement, I will approve and merge.

[TobyRoseman]

Thanks @Ghost-LZW!