Hi Apple Developer Community, I’m exploring ways to fine-tune the SNSoundClassifier to allow users of my iOS app to personalize the model by adding custom sounds or adjusting predictions. While Apple’s WWDC session on sound classification explains how to train from scratch, I’m specifically interested in using SNSoundClassifier as the base model and building/fine-tuning on top of it. Here are a few questions I have: 1. Fine-Tuning on SNSoundClassifier: Is there a way to fine-tune this model programmatically through APIs? The manual approach using macOS, as shown in this documentation is clear, but how can it be done dynamically - within the app for users or in a cloud backend (AWS/iCloud)? Are there APIs or classes that support such on-device/cloud-based fine-tuning or incremental learning? If not directly, can the classifier’s embeddings be used to train a lightweight custom layer? Training is likely computationally intensive and drains too much on battery, doing it on cloud can be right way but need the right apis to get this done. A sample code will do good. 2. Recommended Approach for In-App Model Customization: If SNSoundClassifier doesn’t support fine-tuning, would transfer learning on models like MobileNetV2, YAMNet, OpenL3, or FastViT be more suitable? Given these models (SNSoundClassifier, MobileNetV2, YAMNet, OpenL3, FastViT), which one would be best for accuracy and performance/efficiency on iOS? I aim to maintain real-time performance without sacrificing battery life. Also it is important to see architecture retention and accuracy after conversion to CoreML model. 3. Cost-Effective Backend Setup for Training: Mac EC2 instances on AWS have a 24-hour minimum billing, which can become expensive for limited user requests. Are there better alternatives for deploying and training models on user request when s/he uploads files (training data)? 4. TensorFlow vs PyTorch: Between TensorFlow and PyTorch, which framework would you recommend for iOS Core ML integration? TensorFlow Lite offers mobile-optimized models, but I’m also curious about PyTorch’s performance when converted to Core ML. 5. Metrics: Metrics I have in mind while picking the model are these: Publisher, Accuracy, Fine-Tuning capability, Real-Time/Live use, Suitability of iPhone 16, Architectural retention after coreML conversion, Reasons for unsuitability, Recommended use case. Any insights or recommended approaches would be greatly appreciated. Thanks in advance!

https://developer.apple.com/machine-learning/models/ Adding the DepthAnythingV2SmallF16.mlpackage to a new project in Xcode 16.1 and invoking the class crashes the app. Anyone else having the same issue? I tried Xcode 16.2 beta and it has the same response. Code import UIKit import CoreML class ViewController : UIViewController { override func viewDidLoad() { super.viewDidLoad() // Do any additional setup after loading the view. do { // Use a default model configuration. let defaultConfig = MLModelConfiguration() // app crashes here let model = try? DepthAnythingV2SmallF16( configuration: defaultConfig ) } catch { // } } } Response /AppleInternal/Library/BuildRoots/4b66fb3c-7dd0-11ef-b4fb-4a83e32a47e1/Library/Caches/com.apple.xbs/Sources/MetalPerformanceShadersGraph/mpsgraph/MetalPerformanceShadersGraph/Core/Files/MPSGraphExecutable.mm:129: failed assertion Error: unhandled platform for MPSGraph serialization' `

Hi, I am trying to create a multi label image classifier model using CreateML (the one included in Xcode 16.1). However, my annoations.json file won't get accepted by the app. I get the following error: annotations.json file contains field "Index 0" that is not of type String Here is a JSON example which results in said error: [ { "image": "image1.jpg", "annotations": [ { "label": "car-license-plate", "coordinates": { "x": 160, "y": 108, "width": 190, "height": 200 } } ] }, { "image": "image2.jpg", "annotations": [ { "label": "car-license-plate", "coordinates": { "x": 250, "y": 150, "width": 100, "height": 98 } } ] } ]

Hi Everyone, I'm currently facing an issue where TensorFlow is unable to detect the GPU on my M1 Mac for model training. When I run the following code to check for available GPUs: import tensorflow as tf print("Num GPUs Available: ", len(tf.config.list_physical_devices('GPU'))) Num GPUs Available: 0 I have already applied the steps mentioned in the developer apple document. https://developer.apple.com/metal/tensorflow-plugin/ System Information: Device: M1 Mac Pro Max Python Version: 3.12.2 TensorFlow Version: 2.17.0 OS: macOS Sequoia (15.1) Questions: Is there any additional configuration required to enable GPU support on M1 Macs? Are there specific TensorFlow versions that I should be using for better compatibility? Has anyone else faced this issue, and how did you resolve it?

Hi, I found when continuously predicting with the same Core ML model in 120 FPS will be faster than in 60 FPS. I use Macbook Pro M2 and turn on ProMotion to run Core ML model prediction with a 120 FPS video, the average prediction time is 7.46ms as below: But when I turn off ProMotion, set 60 Hz refresh rate, and run Core ML model prediction with a 60 FPS video, the average prediction time is 10.91ms as below: What could be the technical explanation for these results? Is there any documentation or technical literature that addresses this behavior?

I have followed https://apple.github.io/coremltools/docs-guides/source/installing-coremltools.html but failed. Looks like the doc is too outdated.

In our app we use CoreML. But ever since macOS 15.x was released we started to get a great bunch of crashes like this: Incident Identifier: 424041c3-884b-4e50-bb5a-429a83c3e1c8 CrashReporter Key: B914246B-1291-4D44-984D-EDF84B52310E Hardware Model: Mac14,12 Process: <REMOVED> [1509] Path: /Applications/<REMOVED> Identifier: com.<REMOVED> Version: <REMOVED> Code Type: arm64 Parent Process: launchd [1] Date/Time: 2024-11-13T13:23:06.999Z Launch Time: 2024-11-13T13:22:19Z OS Version: Mac OS X 15.1.0 (24B83) Report Version: 104 Exception Type: SIGABRT Exception Codes: #0 at 0x189042600 Crashed Thread: 36 Thread 36 Crashed: 0 libsystem_kernel.dylib 0x0000000189042600 __pthread_kill + 8 1 libsystem_c.dylib 0x0000000188f87908 abort + 124 2 libsystem_c.dylib 0x0000000188f86c1c __assert_rtn + 280 3 Metal 0x0000000193fdd870 MTLReportFailure.cold.1 + 44 4 Metal 0x0000000193fb9198 MTLReportFailure + 444 5 MetalPerformanceShadersGraph 0x0000000222f78c80 -[MPSGraphExecutable initWithMPSGraphPackageAtURL:compilationDescriptor:] + 296 6 Espresso 0x00000001a290ae3c E5RT::SharedResourceFactory::GetMPSGraphExecutable(std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char> > const&, NSDictionary*) + 932 . . . 43 CoreML 0x0000000192d263bc -[MLModelAsset modelWithConfiguration:error:] + 120 44 CoreML 0x0000000192da96d0 +[MLModel modelWithContentsOfURL:configuration:error:] + 176 45 <REMOVED> 0x000000010497b758 -[<REMOVED> <REMOVED>] (<REMOVED>) No similar crashes on macOS 12-14! MetalPerformanceShadersGraph.log Any clue what is causing this? Thanks! :)

I tried this: struct CarShortcutsProvider: AppShortcutsProvider { @AppShortcutsBuilder static var appShortcuts: [AppShortcut] { AppShortcut( intent: LockCarIntent(), phrases: ["Lock my car with \(.applicationName)", "Lock my \(\.$car) with \(.applicationName)"], shortTitle: LocalizedStringResource("Lock Car"), systemImageName: "lock.fill" ) AppShortcut( intent: UnlockCarIntent(), phrases: ["Unlock my car with \(.applicationName)", "Unlock my \(\.$car) with \(.applicationName)"], shortTitle: LocalizedStringResource("Unlock Car"), systemImageName: "lock.open.fill" ) } } but Siri only understands "unlock my car ", not with the placeholder. Siri asks me then for the car, and it understands it, but not in one sentence. Is there something wrong with my code? Also I tried it without applicationName first, and then it didn't work at all with Siri. Is this a general limitation of app intents? I thought the goal was to reduce friction. If the user has to mention the app name all the time, it adds friction.

Itself been 4-5 days my Image playground has showing the “Downloading Support for Image Playground “

Hey, has anyone figured out how the “Persons” list in Genmoji/Playground actually works? I’ve had a strange experience so far. When I first got access during Beta 2, the list randomly included about 10–15 people, even though my photo library contains many more recognizable faces. To try fixing this, I started naming faces in the Photos app, hoping they’d be added to the Genmoji/Playground list, but nothing changed. Then, after updating to Beta 3, it added just 2–3 of the people I had named. Encouraged, I spent about an hour naming all the faces in my library. But a few hours later, the list unexpectedly removed around 10 people, leaving me with fewer than I had initially. I’ve also read that leaving the phone locked and plugged into power should help sort people in the library, but that hasn’t worked for me yet. Anyone else experienced this or found a way to make it work? Thanks!

We are experiencing a major issue with the native .version1 of the SoundAnalysis framework in iOS 18, which has led to all our user not having recordings. Our core feature relies heavily on sound analysis in the background, and it previously worked flawlessly in prior iOS versions. However, in the new iOS 18, sound analysis stops working in the background, triggering a critical warning. Details of the issue: We are using SoundAnalysis to analyze background sounds and have enabled the necessary background permissions. We are using the latest XCode A warning now appears, and sound analysis fails in the background. Below is the warning message we are encountering: Warning Message: Execution of the command buffer was aborted due to an error during execution. Insufficient Permission (to submit GPU work from background) [Espresso::handle_ex_plan] exception=Espresso exception: "Generic error": Insufficient Permission (to submit GPU work from background) (00000006:kIOGPUCommandBufferCallbackErrorBackgroundExecutionNotPermitted); code=7 status=-1 Unable to compute the prediction using a neural network model. It can be an invalid input data or broken/unsupported model (error code: -1). CoreML prediction failed with Error Domain=com.apple.CoreML Code=0 "Failed to evaluate model 0 in pipeline" UserInfo={NSLocalizedDescription=Failed to evaluate model 0 in pipeline, NSUnderlyingError=0x30330e910 {Error Domain=com.apple.CoreML Code=0 "Failed to evaluate model 1 in pipeline" UserInfo={NSLocalizedDescription=Failed to evaluate model 1 in pipeline, NSUnderlyingError=0x303307840 {Error Domain=com.apple.CoreML Code=0 "Unable to compute the prediction using a neural network model. It can be an invalid input data or broken/unsupported model (error code: -1)." UserInfo={NSLocalizedDescription=Unable to compute the prediction using a neural network model. It can be an invalid input data or broken/unsupported model (error code: -1).}}}}} We urgently need guidance or a fix for this, as our application’s main functionality is severely impacted by this background permission error. Please let us know the next steps or if this is a known issue with iOS 18.

Hi, The most recent version of tensorflow-metal is only available for macosx 12.0 and python up to version 3.11. Is there any chance it could be updated with wheels for macos 15 and Python 3.12 (which is the default version supported for tensrofllow 2.17+)? I'd note that even downgrading to Python 3.11 would not be sufficient, as the wheels only work for macos 12. Thanks.

Hi everyone, I'm working on integrating object recognition from live video feeds into my existing app by following Apple's sample code. My original project captures video and records it successfully. However, after integrating the Vision-based object detection components (VNCoreMLRequest), no detections occur, and the callback for the request is never triggered. To debug this issue, I’ve added the following functionality: Set up AVCaptureVideoDataOutput for processing video frames. Created a VNCoreMLRequest using my Core ML model. The video recording functionality works as expected, but no object detection happens. I’d like to know: How to debug this further? Which key debug points or logs could help identify where the issue lies? Have I missed any key configurations? Below is a diff of the modifications I’ve made to my project for the new feature. Diff of Changes: (Attach the diff provided above) Specific Observations: The captureOutput method is invoked correctly, but there is no output or error from the Vision request callback. Print statements in my setup function setForVideoClassify() show that the setup executes without errors. Questions: Could this be due to issues with my Core ML model compatibility or configuration? Is the VNCoreMLRequest setup incorrect, or do I need to ensure specific image formats for processing? Platform: Xcode 16.1, iOS 18.1, Swift 5, SwiftUI, iPhone 11, Darwin MacBook-Pro.local 24.1.0 Darwin Kernel Version 24.1.0: Thu Oct 10 21:02:27 PDT 2024; root:xnu-11215.41.3~2/RELEASE_X86_64 x86_64 Any guidance or advice is appreciated! Thanks in advance.

I noticed that the ChatGPT is listed as an "Extension" in the Apple Intelligence settings on iOS 18.2 beta. Does this mean developers will be able to create their own extensions? Or will this be limited to larger companies to incorporate their own models into Apple Intelligence?

Hi everyone, Could someone confirm if it's currently possible, or if there are any plans, to restrict users from enabling Apple Intelligence altogether? I understand that we can block individual features using MDM, but I'm interested in knowing if we can prevent users from toggling Apple Intelligence on and off in System Settings entirely. Thanks! Kind Regards, Filipe Nogueira

Hello, I am exploring real-time object detection, and its replacement/overlay with another shape, on live video streams for an iOS app using Core ML and Vision frameworks. My target is to achieve high-speed, real-time detection without noticeable latency, similar to what’s possible with PageFault handling and Associative Caching in OS, but applied to video processing. Given that this requires consistent, real-time model inference, I’m curious about how well the Neural Engine or GPU can handle such tasks on A-series chips in iPhones versus M-series chips (specifically M1 Pro and possibly M4) in MacBooks. Here are a few specific points I’d like insight on: Hardware Suitability: How feasible is it to perform real-time object detection with Core ML on the Neural Engine (i.e., can it maintain low latency)? Would the M-series chips (e.g., M1 Pro or newer) offer a tangible benefit for this type of task compared to the A-series in mobile devices? Which A- and M- chips would be minimum feasible recommendation for such task. Performance Expectations: For continuous, live video object detection, what would be the expected frame rate or latency using an optimized Core ML model? Has anyone benchmarked such applications, and is the M-series required to achieve smooth, real-time processing? Differences Across Apple Hardware: How does performance scale between the A-series Neural Engine and M-series GPU and Neural Engine? Is the M-series vastly superior for real-time Core ML tasks like object detection on live video feeds? If anyone has attempted live object detection on these chips, any insights on real-time performance, limitations, or optimizations would be highly appreciated. Please refer: Apple APIs Thank you in advance for your help!

I have a model that uses a CoreML delegate, and I’m getting the following warning whenever I set the model to nil. My understanding is that CoreML is creating a cache in the app’s storage but is having issues clearing it. As a result, the app’s storage usage increases every time the model is loaded. This StackOverflow post explains the problem in detail: App Storage Size Increases with CoreML usage This is a critical issue because the cache will eventually fill up the phone’s storage: doUnloadModel:options:qos:error:: model=_ANEModel: { modelURL=file:///var/mobile/Containers/Data/Application/22DDB13E-DABA-4195-846F-F884135F37FE/tmp/F38A9824-3944-420C-BD32-78CE598BE22D-10125-00000586EFDFD7D6.mlmodelc/ : sourceURL= (null) : key={"isegment":0,"inputs":{"0_0":{"shape":[256,256,1,3,1]}},"outputs":{"142_0":{"shape":[16,16,1,222,1]},"138_0":{"shape":[16,16,1,111,1]}}} : identifierSource=0 : cacheURLIdentifier=E0CD0F44FB0417936057FC6375770CFDCCC8C698592ED412DDC9C81E96256872_C9D6E5E73302943871DC2C610588FEBFCB1B1D730C63CA5CED15D2CD5A0AC0DA : string_id=0x00000000 : program=_ANEProgramForEvaluation: { programHandle=6077141501305 : intermediateBufferHandle=6077142786285 : queueDepth=127 } : state=3 : programHandle=6077141501305 : intermediateBufferHandle=6077142786285 : queueDepth=127 : attr={ ANEFModelDescription = { ANEFModelInput16KAlignmentArray = ( ); ANEFModelOutput16KAlignmentArray = ( ); ANEFModelProcedures = ( { ANEFModelInputSymbolIndexArray = ( 0 ); ANEFModelOutputSymbolIndexArray = ( 0, 1 ); ANEFModelProcedureID = 0; } ); kANEFModelInputSymbolsArrayKey = ( "0_0" ); kANEFModelOutputSymbolsArrayKey = ( "138_0@output", "142_0@output" ); kANEFModelProcedureNameToIDMapKey = { net = 0; }; }; NetworkStatusList = ( { LiveInputList = ( { BatchStride = 393216; Batches = 1; Channels = 3; Depth = 1; DepthStride = 393216; Height = 256; Interleave = 1; Name = "0_0"; PlaneCount = 3; PlaneStride = 131072; RowStride = 512; Symbol = "0_0"; Type = Float16; Width = 256; } ); LiveOutputList = ( { BatchStride = 113664; Batches = 1; Channels = 111; Depth = 1; DepthStride = 113664; Height = 16; Interleave = 1; Name = "138_0@output"; PlaneCount = 111; PlaneStride = 1024; RowStride = 64; Symbol = "138_0@output"; Type = Float16; Width = 16; }, { BatchStride = 227328; Batches = 1; Channels = 222; Depth = 1; DepthStride = 227328; Height = 16; Interleave = 1; Name = "142_0@output"; PlaneCount = 222; PlaneStride = 1024; RowStride = 64; Symbol = "142_0@output"; Type = Float16; Width = 16; } ); Name = net; } ); } : perfStatsMask=0} was not loaded by the client.

Hello, we're investigating an option to disable writing tools for some customers in our app. I'm aware of the writingToolsBehavior property for UITextView etc, but we would like to have a way to set this globally without having to update all UITextView instances (or future instances). Is there any API to do this? We tried using UITextView.appearance.writingToolsBehavior = .none and it seemed promising on 18.2 beta, however it introduced crashes on devices running 18.1. The crashes look like: *** Terminating app due to uncaught exception 'NSInternalInconsistencyException', reason: 'Have you sent -setWritingToolsBehavior: to <UITextView: 0x14462c000; frame = (0 0; 0 0); text = ''; userInteractionEnabled = NO; gestureRecognizers = <NSArray: 0x30067cb40>; backgroundColor = UIExtendedGrayColorSpace 0 0; layer = <CALayer: 0x3009b1ba0>; contentOffset: {0, 0}; contentSize: {0, 0}; adjustedContentInset: {0, 0, 0, 0}> off the main thread? To verify, look for a complaint in the logs: "Unsupported use of UIKit…", and fix the problem if you find it. If your use is main-thread only please file a radar on UIKit, and attach this log. exercisedImplementations = { "setWritingToolsBehavior:" = ( ); }' Similarly, even on 18.2 beta if we used UITextField.appearance.writingToolsBehavior = .none we would see crashes for any search fields like: *** Terminating app due to uncaught exception 'NSInternalInconsistencyException', reason: 'Have you sent -setWritingToolsBehavior: to <UISearchBarTextField: 0x141c04a00; frame = (0 0; 0 0); text = ''; opaque = NO; gestureRecognizers = <NSArray: 0x301fe15c0>; placeholder = Search Leads; borderStyle = RoundedRect; background = <_UITextFieldSystemBackgroundProvider: 0x3015de960: backgroundView=<_UISearchBarSearchFieldBackgroundView: 0x141c60200; frame = (0 0; 0 0); opaque = NO; autoresize = W+H; userInteractionEnabled = NO; layer = <CALayer: 0x3015de8e0>>, fillColor=(null), textfield=<UISearchBarTextField: 0x141c04a00>>; layer = <CALayer: 0x3015de240>> off the main thread? To verify, look for a complaint in the logs: "Unsupported use of UIKit…", and fix the problem if you find it. If your use is main-thread only please file a radar on UIKit, and attach this log. exercisedImplementations = { "setWritingToolsBehavior:" = ( ); }' Is it possible to set this globally?

I use SoundAnalysis to analyze background sounds and have enabled background permissions. It worked well in previous iOS systems, but a warning appeared in the new iOS18beta version and sound analysis was stopped. Warning List: Execution of the command buffer was aborted due to an error during execution. Insufficient Permission (to submit GPU work from background) [Espresso::handle_ex_plan] exception=Espresso exception: "Generic error": Insufficient Permission (to submit GPU work from background) (00000006:kIOGPUCommandBufferCallbackErrorBackgroundExecutionNotPermitted); code=7 status=-1 Unable to compute the prediction using a neural network model. It can be an invalid input data or broken/unsupported model (error code: -1). CoreML prediction failed with Error Domain=com.apple.CoreML Code=0 "Failed to evaluate model 0 in pipeline" UserInfo={NSLocalizedDescription=Failed to evaluate model 0 in pipeline, NSUnderlyingError=0x30330e910 {Error Domain=com.apple.CoreML Code=0 "Failed to evaluate model 1 in pipeline" UserInfo={NSLocalizedDescription=Failed to evaluate model 1 in pipeline, NSUnderlyingError=0x303307840 {Error Domain=com.apple.CoreML Code=0 "Unable to compute the prediction using a neural network model. It can be an invalid input data or broken/unsupported model (error code: -1)." UserInfo={NSLocalizedDescription=Unable to compute the prediction using a neural network model. It can be an invalid input data or broken/unsupported model (error code: -1).}}}}}

Hi all, When executing an HLO program using the JAX metal PJRT plugin, the program fails due to an unsupported data type returned by the rng_bit_generator operation. The generated HLO includes: %output_state, %output = "mhlo.rng_bit_generator"(%1) <{rng_algorithm = #mhlo.rng_algorithm<PHILOX>}> : (tensor<3xi64>) -> (tensor<3xi64>, tensor<3xui32>) The error message indicates that: Metal only supports MPSDataTypeFloat16, MPSDataTypeBFloat16, MPSDataTypeFloat32, MPSDataTypeInt32, and MPSDataTypeInt64. The use of ui32 seems to be incompatible with Metal’s allowed types. I’m trying to understand if the ui32 output is the problem or maybe the use of rng_bit_generator is wrong. Could you clarify if there is a workaround or planned support for ui32 output in this context? Alternatively, guidance on configuring rng_bit_generator for compatibility with Metal’s supported types would be greatly appreciated.