TimeseriesClassifier crash when update as follows:

"On the latest iOS 18 beta 2, the OCR API，the Translate App and Live Text performs very poorly in recognizing Japanese."

Hi everyone, I attempted to use the MultivariateLinearRegressor from the Create ML Components framework to fit some multi-dimensional data linearly (4 dimensions in my example). I aim to obtain multi-dimensional output points (2 points in my example). However, when I fit the model with my training data and test it, it appears that only the first element of my training data is used for training, regardless of whether I use CreateMLComponents.AnnotatedBatch or [CreateMLComponents.AnnotatedFeature, CoreML.MLShapedArray>] as input. let sourceMatrix: [[Double]] = [ [0,0.1,0.2,0.3], [0.5,0.2,0.6,0.2] ] let referenceMatrix: [[Double]] = [ [0.2,0.7], [0.9,0.1] ] Here is a test code to test the function (ios 18.0 beta, Xcode 16.0 beta) In this example I train the model to learn 2 multidimensional points (4 dimensions) and here are the results of the predictions: ▿ 2 elements ▿ 0 : AnnotatedPrediction<MLShapedArray<Double>, MLShapedArray<Double>> ▿ prediction : 0.20000000298023224 0.699999988079071 ▿ _storage : <StandardStorage<Double>: 0x600002ad8270> ▿ annotation : 0.2 0.7 ▿ _storage : <StandardStorage<Double>: 0x600002b30600> ▿ 1 : AnnotatedPrediction<MLShapedArray<Double>, MLShapedArray<Double>> ▿ prediction : 0.23158159852027893 0.9509953260421753 ▿ _storage : <StandardStorage<Double>: 0x600002ad8c90> ▿ annotation : 0.9 0.1 ▿ _storage : <StandardStorage<Double>: 0x600002b55f20> 0.23158159852027893 0.9509953260421753 is totally random and should be far more closer to [0.9,0.1]. Here is the test code : ( i run it on "My mac, Designed for Ipad") ContentView.swift import CoreImage import CoreImage.CIFilterBuiltins import UIKit import CoreGraphics import Accelerate import Foundation import CoreML import CreateML import CreateMLComponents func createMLShapedArray(from array: [Double], shape: [Int]) -> MLShapedArray<Double> { return MLShapedArray<Double>(scalars: array, shape: shape) } func calculateTransformationMatrixWithNonlinearity(sourceRGB: [[Double]], referenceRGB: [[Double]], degree: Int = 3) async throws -> MultivariateLinearRegressor<Double>.Model { let annotatedFeatures2 = zip(sourceRGB, referenceRGB).map { (featureArray, targetArray) -> AnnotatedFeature<MLShapedArray<Double>, MLShapedArray<Double>> in let featureMLShapedArray = createMLShapedArray(from: featureArray, shape: [featureArray.count]) let targetMLShapedArray = createMLShapedArray(from: targetArray, shape: [targetArray.count]) return AnnotatedFeature(feature: featureMLShapedArray, annotation: targetMLShapedArray) } // Flatten the sourceRGBPoly into a single-dimensional array var flattenedArray = sourceRGB.flatMap { $0 } let featuresMLShapedArray = createMLShapedArray(from: flattenedArray, shape: [2, 4]) flattenedArray = referenceRGB.flatMap { $0 } let targetMLShapedArray = createMLShapedArray(from: flattenedArray, shape: [2, 2]) // Create AnnotatedFeature instances /* let annotatedFeatures2: [AnnotatedFeature<MLShapedArray<Double>, MLShapedArray<Double>>] = [ AnnotatedFeature(feature: featuresMLShapedArray, annotation: targetMLShapedArray) ]*/ let annotatedBatch = AnnotatedBatch(features: featuresMLShapedArray, annotations: targetMLShapedArray) var regressor = MultivariateLinearRegressor<Double>() regressor.configuration.learningRate = 0.1 regressor.configuration.maximumIterationCount=5000 regressor.configuration.batchSize=2 let model = try await regressor.fitted(to: annotatedBatch,validateOn: nil) //var model = try await regressor.fitted(to: annotatedFeatures2) // Proceed to prediction once the model is fitted let predictions = try await model.prediction(from: annotatedFeatures2) // Process or use the predictions print(predictions) print("Predictions:", predictions) return model } struct ContentView: View { var body: some View { VStack {} .onAppear { Task { do { let sourceMatrix: [[Double]] = [ [0,0.1,0.2,0.3], [0.5,0.2,0.6,0.2] ] let referenceMatrix: [[Double]] = [ [0.2,0.7], [0.9,0.1] ] let model = try await calculateTransformationMatrixWithNonlinearity(sourceRGB: sourceMatrix, referenceRGB: referenceMatrix, degree: 2 ) print("Model fitted successfully:", model) } catch { print("Error:", error) } } } } }

I'm getting widespread reports from users trialling iOS 17.6 public beta that Siri Shortcuts are failing whenever they enter any text that looks like a URL. It's getting reported to me because my app happens to have an app intent with a string parameter which can contain a URL in some circumstances. However it's easily reproducible outside of my app: just create a 2 line shortcut like the one below. If you change "This is some text" to "https://www.apple.com" the shortcut below will fail: In iOS 17.5 entering "https://www.apple.com" works fine. I've raised feedback on this (FB14206088) but can anyone confirm that this is indeed a bug and not some weird new feature of Shortcuts where the contents of a variable can somehow change the type of a variable? It would be very, very bad if this were so.

Hi All, I am trying to build a new iOS app by following https://developer.apple.com/videos/play/wwdc2024/10163/?time=67 When I trying to remove all legacy VN I am getting error, I would appreciate if someone can help me get up to speed with the new Vision API

I‘ve created text classification project and selected BERT algorithm With 100 iterations for json file. Json file is valid but training always cancels on 37 iteration… Because tool does not provide any cancellation reasons I have no clue why it happens. Can I check reasons somehow? Or do anyone knows possible reasons or solutions for this?

I have followed the SoupChef example in migrating Custom Intents from SiriKit to AppIntents. However, we only require one iOS release back, so we can require iOS 17. Thus, I eliminated everything that was strictly for backwards compatibility, most notably the SiriKit Extension that required enormous amounts of code to try to coordinate with the real app which worked poorly anyway. I tested for example that an NFC tag Automation created in Shortcuts works to execute an AppIntent while the app is backgrounded. I am now receiving a beta report that indicates someone trying to execute one of our migrated AppIntents from their HomePod is not working, and they say it used to work sometimes (not all the time). I'm sure most such cases used to require the SiriKit Extension in the old SiriKit world. I am terrified that I may need to rebuild that monster once again when the new (to me) AppIntent API seemed so beautiful without it and seemed to work without it. The AppIntent API documentation seems to indicate that SiriKit Extensions are no longer related or required. What is the truth here? Do I need to re-implement everything twice in the SiriKit Extension like a barbarian, or can we live in the new world with AppIntents? Thank you.

Hi, I have a AppEnum and I try to use a custom SF Symbol as DisplayRepresentation.Image but it's not working. I get a blank image when AppEnum picker appears. The custom SF Symbol is stored in the target's asset catalog and it works in the target (eg: Image(named: "custom_sfsymbol"). The issue occurs when I try to use it in a DisplayRepresentation static var caseDisplayRepresentations: [Self: DisplayRepresentation] = [ .sample : DisplayRepresentation(title: "sample_title", image: DisplayRepresentation.Image(named: "custom_sfsymbol")), Can we use a custom SF Symbol in a DisplayRepresentation.Image?

The Keras Embedding layer cannot be calculated on Metal because of the missing Op:StatelessRandomGetKeyCounter, as shown in this error message: tensorflow.python.framework.errors_impl.InvalidArgumentError: Could not satisfy device specification '/job:localhost/replica:0/task:0/device:GPU:0'. enable_soft_placement=0. Supported device types [CPU]. All available devices [/job:localhost/replica:0/task:0/device:GPU:0, /job:localhost/replica:0/task:0/device:CPU:0]. [Op:StatelessRandomGetKeyCounter] A workaround is to enable soft placement, but this obviously is slower: tf.config.set_soft_device_placement(True) Reporting it here as recommended by the TensorFlow Plugin Metal team.

We're using App Intents to launch are control our app via Siri. Siri's responses have been fairly random, some with a "Done" popup, others with a verbal confirmation, others saying "I'm sorry, there's been a problem". The latter is bogus and doesn't look good to potential investors when the app is actually working fine. There appears to be no way in code that I've been able to find so far that would have been tell Siri to STFU. Let us handle our own errors. Otherwise is there a means to supply Siri with a dictionary of restored messages that could be triggered inside the app?

Hi, I am working on creating a EntityPropertyQuery for my App entity. I want the user to be able to use Shortcuts to search by a property in a related entity, but I'm struggling with how the syntax for that looks. I know the documentation for 'EntityPropertyQuery' suggests that this should be possible with a different initializer for the 'QueryProperty' that takes in a 'entityProvider' but I can't figure out how it works. For e.g. my CJPersonAppEntity has 'emails', which is of type CJEmailAppEntity, which has a property 'emailAddress'. I want the user to be able to find the 'person' by looking up an email address. When I try to provide this as a Property to filter by, inside CJPersonAppEntityQuery, but I get a syntax error: static var properties = QueryProperties { Property(\CJPersonEmailAppEntity.$emailAddress, entityProvider: { person in person.emails // error }) { EqualToComparator { NSPredicate(format: "emailAddress == %@", $0) } ContainsComparator { NSPredicate(format: "emailAddress CONTAINS %@", $0) } } } The error says "Cannot convert value of type '[CJPersonEmailAppEntity]' to closure result type 'CJPersonEmailAppEntity'" So it's not expecting an array, but an individual email item. But how do I provide that without running the predicate query that's specified in the closure? So I tried something like this , just returning something without worrying about correctness: Property(\CJPersonEmailAppEntity.$emailAddress, entityProvider: { person in person.emails.first ?? CJPersonEmailAppEntity() // satisfy compiler }) { EqualToComparator { NSPredicate(format: "emailAddress == %@", $0) } ContainsComparator { NSPredicate(format: "emailAddress CONTAINS %@", $0) } } and it built the app, but failed on another the step 'Extracting app intents metadata': error: Entity CJPersonAppEntity does not contain a property named emailAddress. Ensure that the property is wrapped with an @Property property wrapper So I'm not sure what the correct syntax for handling this case is, and I can't find any other examples of how it's done. Would love some feedback for this.

I would like to split up my intents into smaller intents with more atomic pieces of functionality that I can then call one intent from another. For example: struct SumValuesIntent: AppIntent { static var title: LocalizedStringResource { "Sum Values" } let a: Int let b: Int init(a: Int, b: Int) { self.a = a self.b = b } init() { self.init(a: 0, b: 0) } func perform() async throws -> some IntentResult { let sum = a + b print("SumValuesIntent:", sum) return .result(value: sum) } } struct PrintValueIntent: AppIntent { static var title: LocalizedStringResource { "Print Value" } let string: String init(string: String) { self.string = string } init() { self.init(string: "") } func perform() async throws -> some IntentResult { print("PrintValueIntent:", string) return .result() } } What is the best way to chain intents like these? I tried .result(opensIntent: PrintValueIntent(string: String(describing: sum))) as the return type of SumValuesIntent.perform but that doesn't seem to work. Then I tried try await PrintValueIntent(string: String(describing: sum)).perform() as the return type and that works but I'm not sure that's the correct way to do it.

Hi there,I’m a Computer Science student and I have a MacBook Pro 2019 and I’m thinking in buying a new Mac either a Mac Studio or a MacBook Pro but I want to use it for ML. I’m now doing a segmentation model and I’m wondering if I could use Core Ml or the Apple Neural Engine in the new M3 chips to train it, I’m now using colab and tensorflow to create the model but it’s not doing the job, I’m falling short of Cuda memory. Thanks :)

Hi, all. I've been writing various computational functions using Metal. However, in the following operation functions, unlike + and *, there is an accuracy issue in the / operation. This is a function that divides a matrix of shape [n, x, y] and a scalar [1]. When compared to numpy or torch, if I change the operator of the above function to * or + instead of /, I can get completely the same results, but in the case of /, there is a difference in the mean of more than 1e-5. (For reference, this was written with reference to the metal kernel code in llama.cpp) kernel void kernel_div_single_f16( device const half * src0, device const half * src1, device half * dst, constant int64_t & ne00, constant int64_t & ne01, constant int64_t & ne02, constant int64_t & ne03, uint3 tgpig[[threadgroup_position_in_grid]], uint3 tpitg[[thread_position_in_threadgroup]], uint3 ntg[[threads_per_threadgroup]]) { const int64_t i03 = tgpig.z; const int64_t i02 = tgpig.y; const int64_t i01 = tgpig.x; const uint offset = i03*ne02*ne01*ne00 + i02*ne01*ne00 + i01*ne00; for (int i0 = tpitg.x; i0 < ne00; i0 += ntg.x) { dst[offset + i0] = src0[offset+i0] / *src1; } } My mac book is, Macbork Pro(16, 2021) / macOS 12.5 / Apple M1 Pro. Are there any issues related to Div? Thanks in advance for your reply.

Hi, the following model does not run on ANE. Inspecting with deCoreML I see the error ane: Failed to retrieved zero_point. import numpy as np import coremltools as ct from coremltools.converters.mil import Builder as mb import coremltools.converters.mil as mil B, CIN, COUT = 512, 1024, 1024 * 4 @mb.program( input_specs=[ mb.TensorSpec((B, CIN), mil.input_types.types.fp16), ], opset_version=mil.builder.AvailableTarget.iOS18 ) def prog_manual_dequant( x, ): qw = np.random.randint(0, 2 ** 4, size=(COUT, CIN), dtype=np.int8).astype(mil.mil.types.np_uint4_dtype) scale = np.random.randn(COUT, 1).astype(np.float16) offset = np.random.randn(COUT, 1).astype(np.float16) # offset = np.random.randint(0, 2 ** 4, size=(COUT, 1), dtype=np.uint8).astype(mil.mil.types.np_uint4_dtype) dqw = mb.constexpr_blockwise_shift_scale(data=qw, scale=scale, offset=offset) return mb.linear(x=x, weight=dqw) cml_qmodel = ct.convert( prog_manual_dequant, compute_units=ct.ComputeUnit.CPU_AND_NE, compute_precision=ct.precision.FLOAT16, minimum_deployment_target=ct.target.iOS18, ) Whereas if I use an offset with the same dtype as the weights (uint4 in this case), it does run on ANE Tested on coremltools 8.0b1, on macOS 15.0 beta 2/Xcode 15 beta 2, and macOS 15.0 beta 3/Xcode 15 beta 3.

I create an app that allow user tap on button in widget but when app is running it's got an error when tap: Could not find an intent with identifier AppIntentsIdentifier , mangled TypeName: Optional("Somethings")

Like this

How do I directly input landmarks to the activity classifier rather than inputting an image/video?

Hello, I’m currently working on Tiny ML or ML on Edge using the Google Colab platform. Due to the exhaust of my compute unit’s free usage, I’m being prompted to pay. I’ve been considering leveraging the GPU capabilities of my iPad M1 and Intel-based Mac. Both devices utilize Thunderbolt ports capable of sharing connections up to 30GB/s. Since I’m primarily using a classification model, extensive GPU usage isn’t necessary. I’m looking for assistance or guidance on utilizing the iPad’s processor as an eGPU on my Mac, possibly through an API or Apple technology. Any help would be greatly appreciated!

How do I use either of these data sources with MLHandActionClassifierwith on visionOS? MLHandActionClassifier.DataSource.labeledKeypointsDataFrame MLHandActionClassifier.DataSource.labeledKeypointsData visionOS ARKit HandTracking provides us with 27 joints and 3D co-ordinates which differs from the 21 joint, 2D co-ordinates that these two data sources mention in their documentation.