# Mac OS Sonoma not compatible with latest tensorflow-metal libraries **URL:** \ **Category:** TensorFlow\ **Tags:** tensorflow\ **Created:** [September 29, 2023, 2:53am UTC](https://discuss.ai.google.dev/t/mac-os-sonoma-not-compatible-with-latest-tensorflow-metal-libraries/28545 "2023-09-29T02:53:30Z")\ **Posts on this page:** 14\ **Page:** 1
**Author:** ![Erez\_Katz](https://d1dlmcr85iqnpo.cloudfront.net/user_avatar/discuss.ai.google.dev/erez_katz/32/16150_2.png) [@Erez\_Katz](https://discuss.ai.google.dev/u/Erez_Katz)\ **Post date:** [September 29, 2023, 2:53am UTC](https://discuss.ai.google.dev/t/mac-os-sonoma-not-compatible-with-latest-tensorflow-metal-libraries/28545/1 "2023-09-29T02:53:30Z")
I keep seeing output like this – its harmless but massively annoying! I tensorflow/core/grappler/optimizers/custom\_graph\_optimizer\_registry.cc:113] Plugin optimizer for device\_type GPU is enabled. loc(“mps\_select”(“(mpsFileLoc): /AppleInternal/Library/BuildRoots/75428952-3aa4-11ee-8b65-46d450270006/Library/Caches/com.apple.xbs/Sources/MetalPerformanceShadersGraph/mpsgraph/MetalPerformanceShadersGraph/Core/Files/MPSGraphUtilities.mm”:294:0)): error: ‘anec.gain\_offset\_control’ op result #0 must be 4D/5D memref of 16-bit float or 8-bit signed integer or 8-bit unsigned integer values, but got ‘memref\<1x42x1x1xi1\>’ ---
**Author:** ![gautam](https://d1dlmcr85iqnpo.cloudfront.net/user_avatar/discuss.ai.google.dev/gautam/32/16583_2.png) [@gautam](https://discuss.ai.google.dev/u/gautam)\ **Post date:** [September 29, 2023, 3:15am UTC](https://discuss.ai.google.dev/t/mac-os-sonoma-not-compatible-with-latest-tensorflow-metal-libraries/28545/2 "2023-09-29T03:15:17Z")
thanks for sharing this here. much appreciated. Will save me from a nightmare ---
**Author:** ![Renu\_Patel](https://d1dlmcr85iqnpo.cloudfront.net/user_avatar/discuss.ai.google.dev/renu_patel/32/8957_2.png) [@Renu\_Patel](https://discuss.ai.google.dev/u/Renu_Patel)\ **Post date:** [October 5, 2023, 3:02pm UTC](https://discuss.ai.google.dev/t/mac-os-sonoma-not-compatible-with-latest-tensorflow-metal-libraries/28545/4 "2023-10-05T15:02:03Z")
Hi @Erez_Katz Could you please share the minimal reproducible code to replicate the error and to understand the issue? Thank you. ---
**Author:** ![Erez\_Katz](https://d1dlmcr85iqnpo.cloudfront.net/user_avatar/discuss.ai.google.dev/erez_katz/32/16150_2.png) [@Erez\_Katz](https://discuss.ai.google.dev/u/Erez_Katz)\ **Post date:** [October 6, 2023, 2:05am UTC](https://discuss.ai.google.dev/t/mac-os-sonoma-not-compatible-with-latest-tensorflow-metal-libraries/28545/5 "2023-10-06T02:05:55Z")
Thank you Renu – I was able to overcome all the issues with the following (albeit undesirable) solution. I simply disabled the GPU use Seems like the tensorflow macos and tensorflow-metal were incompatible with my other libraries also the GPU code was very slow. I ended up using CPU only and the speed came back and all the errors are gone now too. Apprently GPU is designed for heavy deep models but with simple models, its actually slower. Per macos developer. Here is what I ended up putting in: I set g\_use\_gpu to false naturally to solve the issue if Config.g\_use\_gpu: set\_GPUs() else: # disabling GPU as it is too slow tf.config.set\_visible\_devices(, ‘GPU’) ---
**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 12, 2023, 6:55am UTC](https://discuss.ai.google.dev/t/mac-os-sonoma-not-compatible-with-latest-tensorflow-metal-libraries/28545/7 "2023-10-12T06:55:28Z")
This issue also happen when I use tensorflow 2.9 and metal 0.5 on MacOS Sonoma 14.0 to train a yolov3. I can train it on Ventura ---
**Author:** ![Renu\_Patel](https://d1dlmcr85iqnpo.cloudfront.net/user_avatar/discuss.ai.google.dev/renu_patel/32/8957_2.png) [@Renu\_Patel](https://discuss.ai.google.dev/u/Renu_Patel)\ **Post date:** [October 17, 2023, 5:31am UTC](https://discuss.ai.google.dev/t/mac-os-sonoma-not-compatible-with-latest-tensorflow-metal-libraries/28545/9 "2023-10-17T05:31:01Z")
@Erez_Katz , Okay, Could you please try again by installing the latest `TensorFlow version 2.14` and the compatible `Tensorflow-metal 1.1.0` as mentioned in this [PyPI repository](https://pypi.org/project/tensorflow-metal/) to detect the GPU? Let us know if the issue still persists. Thank you. ---
**Author:** ![Erez\_Katz](https://d1dlmcr85iqnpo.cloudfront.net/user_avatar/discuss.ai.google.dev/erez_katz/32/16150_2.png) [@Erez\_Katz](https://discuss.ai.google.dev/u/Erez_Katz)\ **Post date:** [October 17, 2023, 10:14am UTC](https://discuss.ai.google.dev/t/mac-os-sonoma-not-compatible-with-latest-tensorflow-metal-libraries/28545/10 "2023-10-17T10:14:40Z")
Would you mind sharing the specific command to install both TensorFlow version 2.14 and Tensorflow-metal 1.10 I am asking because my pip install tensorflow==2.14 doesn’t seem to work – it will only recognize 2.13 ---
**Author:** ![Erez\_Katz](https://d1dlmcr85iqnpo.cloudfront.net/user_avatar/discuss.ai.google.dev/erez_katz/32/16150_2.png) [@Erez\_Katz](https://discuss.ai.google.dev/u/Erez_Katz)\ **Post date:** [October 17, 2023, 10:23am UTC](https://discuss.ai.google.dev/t/mac-os-sonoma-not-compatible-with-latest-tensorflow-metal-libraries/28545/11 "2023-10-17T10:23:43Z")
Also - do I still need to use Adam Legacy for my CNN? ---
**Author:** ![pecalvache](https://avatars.discourse-cdn.com/v4/letter/p/848f3c/32.png) [@pecalvache](https://discuss.ai.google.dev/u/pecalvache)\ **Post date:** [October 20, 2023, 4:31am UTC](https://discuss.ai.google.dev/t/mac-os-sonoma-not-compatible-with-latest-tensorflow-metal-libraries/28545/13 "2023-10-20T04:31:43Z")
Hello! I did as you told (I updated the libraries using poetry) and even though I had no errors in my log during the training of a shallow autoencoder with gpu (m2 pro), the speed of training is ridiculously slow! I was getting a 9ms per epoch in Ventura, and now 100ms in Sonoma. I also noticed that the gpu usage does not even get to 15% whereas in Ventura I remember the usage of gpu got up to 70% for the same network. I’m not quite sure if those versions are to be fixed soon or if it is worth to downgrade to Ventura. ---
**Author:** ![Renu\_Patel](https://d1dlmcr85iqnpo.cloudfront.net/user_avatar/discuss.ai.google.dev/renu_patel/32/8957_2.png) [@Renu\_Patel](https://discuss.ai.google.dev/u/Renu_Patel)\ **Post date:** [October 20, 2023, 4:39am UTC](https://discuss.ai.google.dev/t/mac-os-sonoma-not-compatible-with-latest-tensorflow-metal-libraries/28545/14 "2023-10-20T04:39:44Z")
@Erez_Katz , You can use `pip install tensorflow-metal==1.1.0` to install the tensorflow-metal plug-in. Please refer to this [doc](https://developer.apple.com/metal/tensorflow-plugin/) for the same. Thank you. ```auto ``` ---
**Author:** ![Erez\_Katz](https://d1dlmcr85iqnpo.cloudfront.net/user_avatar/discuss.ai.google.dev/erez_katz/32/16150_2.png) [@Erez\_Katz](https://discuss.ai.google.dev/u/Erez_Katz)\ **Post date:** [October 22, 2023, 9:57pm UTC](https://discuss.ai.google.dev/t/mac-os-sonoma-not-compatible-with-latest-tensorflow-metal-libraries/28545/16 "2023-10-22T21:57:48Z")
(ekbase) bridgelineIT@MacBook-Pro-3 ~ % pip install tensorflow-metal==1.1.0 ERROR: Could not find a version that satisfies the requirement tensorflow-metal==1.1.0 (from versions: 0.1.0, 0.1.1, 0.1.2, 0.2.0, 0.3.0, 0.4.0, 0.5.0, 0.5.1, 0.6.0, 0.7.0, 0.7.1, 0.8.0, 1.0.0, 1.0.1) ERROR: No matching distribution found for tensorflow-metal==1.1.0 ---
**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 25, 2023, 3:05am UTC](https://discuss.ai.google.dev/t/mac-os-sonoma-not-compatible-with-latest-tensorflow-metal-libraries/28545/18 "2023-10-25T03:05:37Z")
@Erez_Katz I guess you use python below 3.9. You may need to upgrade python3.9 which could find tensorflow-macos2.14 and 1.1.0. I upgrade to this version and add Adam Legacy for my network. It will not popup this error - ‘anec.gain\_offset\_control’ op. When I train a yolov3, I suffer an error called as below with M2 /AppleInternal/Library/BuildRoots/90c9c1ae-37b6-11ee-a991-46d450270006/Library/Caches/com.apple.xbs/Sources/MetalPerformanceShaders/MPSCore/Utility/MPSLibrary.mm:550: failed assertion `MPSKernel MTLComputePipelineStateCache unable to load function ndArrayConvolution2DGradientWithWeightsA14. Compute function exceeds available temporary registers: (null) I try to set environment veriables - MTL\_SHADER\_VALIDATION=1. It can train, but I found the speed of training is 5-7 times slower than upgrade Sonoma. I also monitor that the GPU is using. I do not know why the speed is so slower. It seem to related the - MTL\_SHADER\_VALIDATION=1. But if i am not set this, it can not train failed with GPU ---
**Author:** ![Tom\_Xu](https://d1dlmcr85iqnpo.cloudfront.net/user_avatar/discuss.ai.google.dev/tom_xu/32/17205_2.png) [@Tom\_Xu](https://discuss.ai.google.dev/u/Tom_Xu)\ **Post date:** [October 26, 2023, 4:07am UTC](https://discuss.ai.google.dev/t/mac-os-sonoma-not-compatible-with-latest-tensorflow-metal-libraries/28545/19 "2023-10-26T04:07:26Z")
I am using python 3.9 and tf-macos2.14 and try to run a deep learning keras based model I used to be able to run for tf2.11-macos. it can run but I got so much annoying warning like the following: loc(“mps\_select”(“(mpsFileLoc): [/AppleInternal/Library/BuildRoots/75428952-3aa4-11ee-8b65-46d450270006/Library/Caches/com.apple.xbs/Sources/MetalPerformanceShadersGraph/mpsgraph/MetalPerformanceShadersGraph/Core/Files/MPSGraphUtilities.mm](https://file+.vscode-resource.vscode-cdn.net/AppleInternal/Library/BuildRoots/75428952-3aa4-11ee-8b65-46d450270006/Library/Caches/com.apple.xbs/Sources/MetalPerformanceShadersGraph/mpsgraph/MetalPerformanceShadersGraph/Core/Files/MPSGraphUtilities.mm)”:294:0)): error: ‘anec.gain\_offset\_control’ op result #0 must be 4D/5D memref of 16-bit float or 8-bit signed integer or 8-bit unsigned integer values, but got ‘memref\<1x2063x1x20xi1\>’ ---
**Author:** ![Tuukka\_Nieminen](https://d1dlmcr85iqnpo.cloudfront.net/user_avatar/discuss.ai.google.dev/tuukka_nieminen/32/6891_2.png) [@Tuukka\_Nieminen](https://discuss.ai.google.dev/u/Tuukka_Nieminen)\ **Post date:** [November 17, 2023, 9:18am UTC](https://discuss.ai.google.dev/t/mac-os-sonoma-not-compatible-with-latest-tensorflow-metal-libraries/28545/21 "2023-11-17T09:18:54Z")
I use macOS Sonoma version 14.1 and Apple M2 Pro. Regarding the combo python 3.9, tensorflow 2.14.0 (or 2.14.1) and tensorflow-metal 1.1.0, the training proceeds as follows: Epoch 1/50 431/431 [==============================] - 166s 222ms/step - loss: 0.2384 - sparse\_categorical\_accuracy: 0.9313 - val\_loss: 0.2923 - val\_sparse\_categorical\_accuracy: 0.9290 Epoch 2/50 431/431 [==============================] - 136s 315ms/step - loss: 0.1056 - sparse\_categorical\_accuracy: 0.9671 - val\_loss: 0.1287 - val\_sparse\_categorical\_accuracy: 0.9568 Epoch 3/50 431/431 [==============================] - 1062s 2s/step - loss: 0.0835 - sparse\_categorical\_accuracy: 0.9719 - val\_loss: 0.1326 - val\_sparse\_categorical\_accuracy: 0.9596 Epoch 4/50 431/431 [==============================] - 419s 972ms/step - loss: 0.0863 - sparse\_categorical\_accuracy: 0.9721 - val\_loss: 0.1155 - val\_sparse\_categorical\_accuracy: 0.9629 Epoch 5/50 431/431 [==============================] - 166s 383ms/step - loss: 0.1069 - sparse\_categorical\_accuracy: 0.9691 - val\_loss: 0.1385 - val\_sparse\_categorical\_accuracy: 0.9555 In a nutshell, the training is ridiculously slow and the loss does not really decrease, but either increases of stagnates to a really bad level. If I use python 3.9, tensorflow 2.13.0 and tensorflow-metal 0.5.1 instead, the training is fast, and proceeds as expected: Epoch 1/50 431/431 [==============================] - 16s 34ms/step - loss: 0.1884 - sparse\_categorical\_accuracy: 0.9306 - val\_loss: 0.0764 - val\_sparse\_categorical\_accuracy: 0.9668 Epoch 2/50 431/431 [==============================] - 16s 34ms/step - loss: 0.0665 - sparse\_categorical\_accuracy: 0.9746 - val\_loss: 0.0529 - val\_sparse\_categorical\_accuracy: 0.9790 Epoch 3/50 431/431 [==============================] - 16s 35ms/step - loss: 0.0517 - sparse\_categorical\_accuracy: 0.9798 - val\_loss: 0.0516 - val\_sparse\_categorical\_accuracy: 0.9795 Epoch 4/50 431/431 [==============================] - 16s 35ms/step - loss: 0.0453 - sparse\_categorical\_accuracy: 0.9825 - val\_loss: 0.0397 - val\_sparse\_categorical\_accuracy: 0.9842 Epoch 5/50 431/431 [==============================] - 16s 35ms/step - loss: 0.0418 - sparse\_categorical\_accuracy: 0.9839 - val\_loss: 0.0372 - val\_sparse\_categorical\_accuracy: 0.9858 but the terminal gets filled by the following error messages: loc(“mps\_select”(“(mpsFileLoc): /AppleInternal/Library/BuildRoots/495c257e-668e-11ee-93ce-926038f30c31/Library/Caches/com.apple.xbs/Sources/MetalPerformanceShadersGraph/mpsgraph/MetalPerformanceShadersGraph/Core/Files/MPSGraphUtilities.mm”:294:0)): error: ‘anec.gain\_offset\_control’ op result #0 must be 4D/5D memref of 16-bit float or 8-bit signed integer or 8-bit unsigned integer values, but got ‘memref\<1x1x1x1xi1\>’