fix(wordspotting): zero CTC keyword-spotter audio padding

Sources/FluidAudio/ASR/Parakeet/SlidingWindow/CustomVocabulary/WordSpotting/CtcKeywordSpotter+Inference.swift

@@ -234,8 +234,6 @@ extension CtcKeywordSpotter {
     // MARK: - Audio Preparation
 
     private func prepareAudioArray(_ audioSamples: [Float]) throws -> (MLMultiArray, Int) {
-        let clampedCount = min(audioSamples.count, maxModelSamples)
-
         // Detect expected input rank from the MelSpectrogram model's 'audio' feature description.
         // Canary-1b-v2 expects rank 1 [samples], parakeet-ctc-0.6b expects rank 2 [1, samples].
         let melModel = models.melSpectrogram
@@ -246,19 +244,12 @@ extension CtcKeywordSpotter {
         let dataType: MLMultiArrayDataType =
             audioDesc?.multiArrayConstraint?.dataType == .float16 ? .float16 : .float32
 
-        let array: MLMultiArray
-        if expectedRank == 2 {
-            // Rank 2: [1, maxSamples]
-            array = try MLMultiArray(shape: [1, NSNumber(value: maxModelSamples)], dataType: dataType)
-        } else {
-            // Rank 1: [maxSamples]
-            array = try MLMultiArray(shape: [NSNumber(value: maxModelSamples)], dataType: dataType)
-        }
-
-        // Copy actual samples (MLMultiArray is zero-initialized, so padding is implicit).
-        for i in 0..<clampedCount {
-            array[i] = NSNumber(value: audioSamples[i])
-        }
+        let (array, clampedCount) = try Self.makePaddedAudioArray(
+            audioSamples,
+            maxSamples: maxModelSamples,
+            rank: expectedRank,
+            dataType: dataType
+        )
 
         if debugMode {
             let midpoint = clampedCount / 2
@@ -278,6 +269,36 @@ extension CtcKeywordSpotter {
         return (array, clampedCount)
     }
 
+    /// Pad `audioSamples` to `maxSamples` and zero the unused tail.
+    ///
+    /// `MLMultiArray(shape:dataType:)` does not zero storage. The mel model reads
+    /// the whole window, so the tail is cleared before the clip is copied.
+    static func makePaddedAudioArray(
+        _ audioSamples: [Float],
+        maxSamples: Int,
+        rank: Int,
+        dataType: MLMultiArrayDataType
+    ) throws -> (MLMultiArray, Int) {
+        let clampedCount = min(audioSamples.count, maxSamples)
+
+        let array: MLMultiArray
+        if rank == 2 {
+            // Rank 2: [1, maxSamples]
+            array = try MLMultiArray(shape: [1, NSNumber(value: maxSamples)], dataType: dataType)
+        } else {
+            // Rank 1: [maxSamples]
+            array = try MLMultiArray(shape: [NSNumber(value: maxSamples)], dataType: dataType)
+        }
+
+        // The shape initializer leaves storage uninitialized. Clear it, then copy the clip.
+        array.reset(to: 0)
+        for i in 0..<clampedCount {
+            array[i] = NSNumber(value: audioSamples[i])
+        }
+
+        return (array, clampedCount)
+    }
+
     private func makeAudioFeatureProvider(array: MLMultiArray, length: Int) throws -> MLFeatureProvider {
         let lengthArray = try MLMultiArray(shape: [1], dataType: .int32)
         lengthArray[0] = NSNumber(value: length)

Tests/FluidAudioTests/ASR/Parakeet/SlidingWindow/CustomVocabulary/CtcKeywordSpotterAudioPaddingTests.swift

@@ -0,0 +1,99 @@
+import CoreML
+import XCTest
+
+@testable import FluidAudio
+
+/// `CtcKeywordSpotter.makePaddedAudioArray` must zero the tail past the copied clip.
+/// `MLMultiArray(shape:dataType:)` does not, and the mel model reads that tail.
+final class CtcKeywordSpotterAudioPaddingTests: XCTestCase {
+
+    // MARK: - Short clip
+
+    func testShortClipFloat32ZerosTheTail() throws {
+        let maxSamples = 32
+        let samples: [Float] = [0.25, -3, 2]
+        XCTAssertTrue(samples[0].isFinite && samples[0] != 0)
+
+        // Fresh allocations: a recycled dirty page shows up when the tail is left uninitialized.
+        for _ in 0..<60 {
+            for rank in [1, 2] {
+                try assertPaddedAudio(samples, maxSamples: maxSamples, rank: rank, dataType: .float32)
+            }
+        }
+    }
+
+    /// Same contract for `.float16`. Values are exactly representable, and reads use
+    /// `NSNumber.floatValue` only (macOS x86_64 has no Swift `Float16`).
+    func testShortClipFloat16ZerosTheTail() throws {
+        let maxSamples = 24
+        let samples: [Float] = [1, 0.5, -1.25]
+        XCTAssertTrue(samples[0].isFinite && samples[0] != 0)
+
+        for _ in 0..<60 {
+            for rank in [1, 2] {
+                try assertPaddedAudio(samples, maxSamples: maxSamples, rank: rank, dataType: .float16)
+            }
+        }
+    }
+
+    // MARK: - Window boundaries
+
+    func testInputMatchingWindowHasNoPadding() throws {
+        let maxSamples = 16
+        let samples = (0..<maxSamples).map { Float($0) + 0.5 }
+        for rank in [1, 2] {
+            try assertPaddedAudio(samples, maxSamples: maxSamples, rank: rank, dataType: .float32)
+        }
+    }
+
+    func testInputLongerThanWindowKeepsLeadingPrefix() throws {
+        let maxSamples = 16
+        let samples = (0..<(maxSamples + 8)).map { Float($0) + 0.5 }
+        for rank in [1, 2] {
+            try assertPaddedAudio(samples, maxSamples: maxSamples, rank: rank, dataType: .float32)
+        }
+    }
+
+    func testEmptyInputIsAllZeros() throws {
+        let maxSamples = 20
+        for rank in [1, 2] {
+            try assertPaddedAudio([], maxSamples: maxSamples, rank: rank, dataType: .float32)
+        }
+    }
+
+    // MARK: - Contract
+
+    private func assertPaddedAudio(
+        _ samples: [Float],
+        maxSamples: Int,
+        rank: Int,
+        dataType: MLMultiArrayDataType,
+        file: StaticString = #filePath,
+        line: UInt = #line
+    ) throws {
+        let (array, clampedCount) = try CtcKeywordSpotter.makePaddedAudioArray(
+            samples,
+            maxSamples: maxSamples,
+            rank: rank,
+            dataType: dataType
+        )
+        let copiedCount = min(samples.count, maxSamples)
+
+        XCTAssertEqual(clampedCount, copiedCount, file: file, line: line)
+        XCTAssertEqual(array.count, maxSamples, file: file, line: line)
+        XCTAssertEqual(array.dataType, dataType, file: file, line: line)
+        let expectedShape = rank == 2 ? [1, maxSamples] : [maxSamples]
+        XCTAssertEqual(array.shape.map(\.intValue), expectedShape, file: file, line: line)
+
+        for index in 0..<array.count {
+            let value = array[index].floatValue
+            if index < clampedCount {
+                XCTAssertEqual(value, samples[index], file: file, line: line)
+                XCTAssertTrue(value.isFinite, file: file, line: line)
+            } else {
+                XCTAssertTrue(value.isFinite, "padding \(index) should be finite", file: file, line: line)
+                XCTAssertEqual(value, Float(0), "padding \(index) should be 0", file: file, line: line)
+            }
+        }
+    }
+}