# MacOS M2 upgrade Sonoma 14.0 can not train model with tensorflow and karas

**URL:** <https://discuss.ai.google.dev/t/macos-m2-upgrade-sonoma-14-0-can-not-train-model-with-tensorflow-and-karas/32287>\
**Category:** TFX-Addons\
**Tags:** mac\_os\
**Created:** [October 13, 2023, 12:45pm UTC](https://discuss.ai.google.dev/t/macos-m2-upgrade-sonoma-14-0-can-not-train-model-with-tensorflow-and-karas/32287 "2023-10-13T12:45:08Z")\
**Posts on this page:** 2\
**Page:** 1

<div class="post-metadata">

**Author:** ![Cloris\_Yin](https://d1dlmcr85iqnpo.cloudfront.net/user_avatar/discuss.ai.google.dev/cloris_yin/32/17719_2.png) [@Cloris\_Yin](https://discuss.ai.google.dev/u/Cloris_Yin)\
**Post date:** [October 13, 2023, 12:45pm UTC](https://discuss.ai.google.dev/t/macos-m2-upgrade-sonoma-14-0-can-not-train-model-with-tensorflow-and-karas/32287/1 "2023-10-13T12:45:08Z")

</div>

I can train a yolov3 at MacOS M2 ventura with tensorflow-macos=2.9.0(karas=2.9.0) and tensorflow-mental=0.5.  
But when I upgrade the system to Sonoma14.0. I can not train model with below error.’

`MPSKernel MTLComputePipelineStateCache unable to load function ndArrayConvolution2DGradientWithWeightsA14. Compute function exceeds available temporary registers: (null)

I also try upgrade the tensorflow-mac to 2.13 and metal to 1.0.0. The same code on M1 somoma system can work well. This is only happen on M2. Is there any one know the this?

---

<div class="post-metadata">

**Author:** ![Divya\_Sree\_Kayyuri](https://d1dlmcr85iqnpo.cloudfront.net/user_avatar/discuss.ai.google.dev/divya_sree_kayyuri/32/29791_2.png) [@Divya\_Sree\_Kayyuri](https://discuss.ai.google.dev/u/Divya_Sree_Kayyuri)\
**Post date:** [July 29, 2025, 7:39am UTC](https://discuss.ai.google.dev/t/macos-m2-upgrade-sonoma-14-0-can-not-train-model-with-tensorflow-and-karas/32287/3 "2025-07-29T07:39:49Z")

</div>

Hi @Cloris_Yin,  
Apologies for the delayed response.  
Please try using the latest versions of `TensorFlow` and `TensorFlow-Metal` . Kindly, refer to this [document](https://pypi.org/project/tensorflow-metal/1.2.0/) for compatible versions. Thank you!