Auto-format via swift-format (#870)

Co-authored-by: Dan Zheng <danielzheng@google.com>

Author: asuhan

Package.swift

@@ -20,7 +20,7 @@ import PackageDescription
 let package = Package(
   name: "TensorFlow",
   platforms: [
-    .macOS(.v10_13),
+    .macOS(.v10_13)
   ],
   products: [
     .library(

Sources/TensorFlow/Core/MixedPrecision.swift

@@ -157,7 +157,7 @@ extension Tensor {
   }
 
   /// Promotes a scalar to a tensor with the same device and precision as the given tensor.
-  @differentiable( where Scalar: TensorFlowFloatingPoint)
+  @differentiable(where Scalar: TensorFlowFloatingPoint)
   public init(_ value: Scalar, deviceAndPrecisionLike tensor: Tensor) {
     let device = tensor.device
     let tmp = Tensor(value, on: device)

Sources/TensorFlow/Core/Runtime.swift

@@ -347,33 +347,33 @@ extension _ExecutionContext {
   }
 }
 
-  internal func _trace<In: TensorGroup, Out: TensorGroup>(_ fn: (In) -> Out) -> TFFunction {
-    let useLazyTensor = _ThreadLocalState.useLazyTensor
-    defer { _ThreadLocalState.useLazyTensor = useLazyTensor }
-    _ThreadLocalState.useLazyTensor = true
-    let trace = LazyTensorTraceBuilder.trace(fn)
-    return TFFunction(trace: trace)
-  }
+internal func _trace<In: TensorGroup, Out: TensorGroup>(_ fn: (In) -> Out) -> TFFunction {
+  let useLazyTensor = _ThreadLocalState.useLazyTensor
+  defer { _ThreadLocalState.useLazyTensor = useLazyTensor }
+  _ThreadLocalState.useLazyTensor = true
+  let trace = LazyTensorTraceBuilder.trace(fn)
+  return TFFunction(trace: trace)
+}
 
-  // Trace the given function to generate a TF graph and return a closure that can be used to launch
-  // the graph.
-  public func _graph<In: TensorGroup, Out: TensorGroup>(
-    _ fn: (In) -> Out,
-    useXLA: Bool = false
-  ) -> (In) -> Out {
-    let tffunc = _trace(fn)
-    return { input in
-      let inputHandles = input._tensorHandles.map { $0._tfeTensorHandle }
-      let outputHandles = tffunc.execute(inputHandles, usingXLA: useXLA)
-      return Out(_handles: outputHandles)
-    }
+// Trace the given function to generate a TF graph and return a closure that can be used to launch
+// the graph.
+public func _graph<In: TensorGroup, Out: TensorGroup>(
+  _ fn: (In) -> Out,
+  useXLA: Bool = false
+) -> (In) -> Out {
+  let tffunc = _trace(fn)
+  return { input in
+    let inputHandles = input._tensorHandles.map { $0._tfeTensorHandle }
+    let outputHandles = tffunc.execute(inputHandles, usingXLA: useXLA)
+    return Out(_handles: outputHandles)
   }
+}
 
 /// Trace the given function and return the name of the corresponding `TF_Function: In -> Out` that
 /// was created.
 public func _tffunc<In: TensorGroup, Out: TensorGroup>(_ fn: (In) -> Out) -> String {
-    let tffunc = _trace(fn)
-    return tffunc.name
+  let tffunc = _trace(fn)
+  return tffunc.name
 }
 
 extension _ExecutionContext {

Sources/TensorFlow/Core/Tensor.swift

@@ -131,7 +131,7 @@ extension Tensor {
   /// Reshape to scalar.
   /// - Precondition: The tensor has exactly one scalar.
   @inlinable
-  @differentiable( where Scalar: TensorFlowFloatingPoint)
+  @differentiable(where Scalar: TensorFlowFloatingPoint)
   public func scalarized() -> Scalar {
     precondition(
       shape.contiguousSize == 1,
@@ -174,7 +174,7 @@ extension Tensor {
     return handle.makeHostCopy()
   }
 
-  @differentiable( where Scalar: TensorFlowFloatingPoint)
+  @differentiable(where Scalar: TensorFlowFloatingPoint)
   public var scalars: [Scalar] {
     #if USING_X10_BACKEND
       if handle.backend == .XLA {
@@ -205,7 +205,7 @@ extension Tensor where Scalar: TensorFlowFloatingPoint {
 
 extension Tensor {
   /// Creates a 0-D tensor from a scalar value.
-  @differentiable( where Scalar: TensorFlowFloatingPoint)
+  @differentiable(where Scalar: TensorFlowFloatingPoint)
   public init(_ value: Scalar, on device: Device = .default) {
     #if USING_X10_BACKEND
       switch device.backend {
@@ -233,7 +233,7 @@ extension Tensor where Scalar: TensorFlowFloatingPoint {
 extension Tensor {
   /// Creates a 1D tensor from scalars.
   @inlinable
-  @differentiable( where Scalar: TensorFlowFloatingPoint)
+  @differentiable(where Scalar: TensorFlowFloatingPoint)
   public init(_ scalars: [Scalar], on device: Device = .default) {
     self.init(shape: [scalars.count], scalars: scalars, on: device)
   }
@@ -266,7 +266,7 @@ extension Tensor {
   ///   - scalars: The scalar contents of the tensor.
   /// - Precondition: The product of the dimensions of the shape must equal the number of scalars.
   @inlinable
-  @differentiable( where Scalar: TensorFlowFloatingPoint)
+  @differentiable(where Scalar: TensorFlowFloatingPoint)
   public init(shape: TensorShape, scalars: [Scalar], on device: Device = .default) {
     precondition(
       shape.contiguousSize == scalars.count,
@@ -677,15 +677,15 @@ extension Tensor: AdditiveArithmetic where Scalar: Numeric {
   /// Adds two tensors and produces their sum.
   /// - Note: `+` supports broadcasting.
   @inlinable
-  @differentiable( where Scalar: TensorFlowFloatingPoint)
+  @differentiable(where Scalar: TensorFlowFloatingPoint)
   public static func + (lhs: Tensor, rhs: Tensor) -> Tensor {
     _Raw.addV2(lhs, rhs)
   }
 
   /// Subtracts one tensor from another and produces their difference.
   /// - Note: `-` supports broadcasting.
   @inlinable
-  @differentiable( where Scalar: TensorFlowFloatingPoint)
+  @differentiable(where Scalar: TensorFlowFloatingPoint)
   public static func - (lhs: Tensor, rhs: Tensor) -> Tensor {
     _Raw.sub(lhs, rhs)
   }

Sources/TensorFlow/Core/TensorHandle.swift

@@ -82,7 +82,7 @@ public class TFETensorHandle: _AnyTensorHandle {
     }
   }
 
-    public var backend: Device.Backend { .TF_EAGER }
+  public var backend: Device.Backend { .TF_EAGER }
 }
 
 /// `TensorHandle` is the type used by ops. It includes a `Scalar` type, which
@@ -173,12 +173,12 @@ extension TensorHandle {
     get { handle.shape }
   }
 
-    /// The backend used to dispatch ops.
-    @inlinable
-    public var backend: Device.Backend {
-      @_semantics("autodiff.nonvarying")
-      get { handle.backend }
-    }
+  /// The backend used to dispatch ops.
+  @inlinable
+  public var backend: Device.Backend {
+    @_semantics("autodiff.nonvarying")
+    get { handle.backend }
+  }
 }
 
 extension TensorHandle {

Sources/TensorFlow/Initializers.swift

@@ -34,7 +34,7 @@ extension Tensor {
   ///   - repeatedValue: The scalar value to repeat.
   ///   - shape: The dimensions of the tensor.
   @inlinable
-  @differentiable( where Scalar: TensorFlowFloatingPoint)
+  @differentiable(where Scalar: TensorFlowFloatingPoint)
   public init(
     repeating repeatedValue: Scalar, shape: TensorShape,
     on device: Device = .default
@@ -91,7 +91,7 @@ extension Tensor where Scalar: Numeric {
 
   /// Perform an element-wise conversion from another `Tensor`.
   @inlinable
-  @differentiable( where Scalar: TensorFlowFloatingPoint, OtherScalar: TensorFlowFloatingPoint)
+  @differentiable(where Scalar: TensorFlowFloatingPoint, OtherScalar: TensorFlowFloatingPoint)
   public init<OtherScalar: Numeric>(_ other: Tensor<OtherScalar>) {
     self = _Raw.cast(other)
   }
@@ -114,7 +114,7 @@ extension Tensor where Scalar: TensorFlowFloatingPoint {
 extension Tensor {
   /// Creates a tensor from an array of tensors (which may themselves be scalars).
   @inlinable
-  @differentiable( where Scalar: TensorFlowFloatingPoint)
+  @differentiable(where Scalar: TensorFlowFloatingPoint)
   public init(_ elements: [Tensor]) {
     self = _Raw.pack(elements)
   }
@@ -148,7 +148,7 @@ extension Tensor {
   ///
   /// - Returns: The stacked tensor.
   @inlinable
-  @differentiable( where Scalar: TensorFlowFloatingPoint)
+  @differentiable(where Scalar: TensorFlowFloatingPoint)
   public init(stacking tensors: [Tensor], alongAxis axis: Int = 0) {
     self = _Raw.pack(tensors, axis: Int64(axis))
   }
@@ -186,7 +186,7 @@ extension Tensor {
   ///
   /// - Returns: The concatenated tensor.
   @inlinable
-  @differentiable( where Scalar: TensorFlowFloatingPoint)
+  @differentiable(where Scalar: TensorFlowFloatingPoint)
   public init(concatenating tensors: [Tensor], alongAxis axis: Int = 0) {
     precondition(tensors.count > 0)
     self = _Raw.concatV2(tensors, axis: Tensor<Int32>(Int32(axis), on: tensors.first!.device))

Sources/TensorFlow/Layers/Dropout.swift

@@ -85,7 +85,7 @@ public struct Dropout<Scalar: TensorFlowFloatingPoint>: ParameterlessLayer {
 @frozen
 public struct AlphaDropout<Scalar: TensorFlowFloatingPoint>: ParameterlessLayer {
   @noDerivative public let probability: Double
-  
+
   /// Initializes an `AlphaDropout` layer with a configurable `probability`.
   ///
   /// - Parameter probability: The probability of a node dropping out.
@@ -113,11 +113,11 @@ public struct AlphaDropout<Scalar: TensorFlowFloatingPoint>: ParameterlessLayer
       let b = -a * alpha_p * probability
 
       // Apply mask
-      var x = input * Tensor(noise) 
+      var x = input * Tensor(noise)
       x = x + Scalar(alpha_p) * (1 - Tensor(noise))
 
       // Do affine transformation
-      return Scalar(a) * x  + Scalar(b)
+      return Scalar(a) * x + Scalar(b)
     case .inference:
       return input
     }

Sources/TensorFlow/Layers/Pooling.swift

@@ -444,8 +444,10 @@ public struct FractionalMaxPool2D<Scalar: TensorFlowFloatingPoint>: Parameterles
   /// A second seed to avoid seed collision
   @noDerivative public let seed2: Int64
   /// Initializes a `FractionalMaxPool` layer with configurable `poolingRatio`.
-  public init(poolingRatio: (Double, Double, Double, Double), pseudoRandom: Bool = false,
-    overlapping: Bool = false, deterministic: Bool = false, seed: Int64 = 0, seed2: Int64 = 0) {
+  public init(
+    poolingRatio: (Double, Double, Double, Double), pseudoRandom: Bool = false,
+    overlapping: Bool = false, deterministic: Bool = false, seed: Int64 = 0, seed2: Int64 = 0
+  ) {
     precondition(
       poolingRatio.0 == 1.0 && poolingRatio.3 == 1.0,
       "Pooling on batch and channels dimensions not supported.")
@@ -468,7 +470,7 @@ public struct FractionalMaxPool2D<Scalar: TensorFlowFloatingPoint>: Parameterles
   public func callAsFunction(_ input: Tensor<Scalar>) -> Tensor<Scalar> {
     fractionalMaxPool2D(
       input,
-      poolingRatio: poolingRatio, 
+      poolingRatio: poolingRatio,
       pseudoRandom: pseudoRandom,
       overlapping: overlapping,
       deterministic: deterministic,
@@ -490,8 +492,10 @@ extension FractionalMaxPool2D {
   ///     used when iterating over a FractionalMaxPool2D node in the computation graph.
   ///   - seed: A seed for random number generator.
   ///   - seed2: A second seed to avoid seed collision.
-  public init(poolingRatio: (Double, Double), pseudoRandom: Bool = false,
-    overlapping: Bool = false, deterministic: Bool = false, seed: Int64 = 0, seed2: Int64 = 0) {
+  public init(
+    poolingRatio: (Double, Double), pseudoRandom: Bool = false,
+    overlapping: Bool = false, deterministic: Bool = false, seed: Int64 = 0, seed2: Int64 = 0
+  ) {
     self.init(
       poolingRatio: (1.0, poolingRatio.0, poolingRatio.1, 1.0),
       pseudoRandom: pseudoRandom,

Sources/TensorFlow/Layers/Recurrent.swift

@@ -369,7 +369,7 @@ public struct RecurrentLayer<Cell: RecurrentLayerCell>: Layer {
     return timeStepOutputs
   }
 
-  @differentiable(wrt: (self,inputs,initialState))
+  @differentiable(wrt: (self, inputs, initialState))
   public func call(
     _ inputs: [Cell.TimeStepInput],
     initialState: Cell.State

Sources/TensorFlow/Operators/Basic.swift

@@ -16,7 +16,7 @@ infix operator .!=: ComparisonPrecedence
 
 /// Returns a tensor with the same shape and scalars as the specified tensor.
 @inlinable
-@differentiable( where Scalar: TensorFlowFloatingPoint)
+@differentiable(where Scalar: TensorFlowFloatingPoint)
 public func identity<Scalar>(_ x: Tensor<Scalar>) -> Tensor<Scalar> {
   x
 }
@@ -57,7 +57,7 @@ extension Tensor {
   ///
   /// - Returns: Array containing the unstacked tensors.
   @inlinable
-  @differentiable( where Scalar: TensorFlowFloatingPoint)
+  @differentiable(where Scalar: TensorFlowFloatingPoint)
   public func unstacked(alongAxis axis: Int = 0) -> [Tensor] {
     ensureValid(axis: axis)
     let posAxis = axis < 0 ? axis + rank : axis
@@ -87,7 +87,7 @@ extension Tensor {
   ///
   /// - Returns: An array containing the tensors part.
   @inlinable
-  @differentiable( where Scalar: TensorFlowFloatingPoint)
+  @differentiable(where Scalar: TensorFlowFloatingPoint)
   public func split(count: Int, alongAxis axis: Int = 0) -> [Tensor] {
     ensureValid(axis: axis)
     precondition(
@@ -429,7 +429,7 @@ extension Tensor {
   ///   specified axis.
   /// - Precondition: The axis must be in the range `-rank..<rank`.
   @inlinable
-  @differentiable( where Scalar: TensorFlowFloatingPoint)
+  @differentiable(where Scalar: TensorFlowFloatingPoint)
   public func concatenated(with other: Tensor, alongAxis axis: Int = 0) -> Tensor {
     return Tensor(concatenating: [self, other], alongAxis: axis)
   }
@@ -440,7 +440,7 @@ extension Tensor {
   ///   and may be controversial. The existence/naming of `++` will be discussed
   ///   during a later API design phase.
   @inlinable
-  @differentiable( where Scalar: TensorFlowFloatingPoint)
+  @differentiable(where Scalar: TensorFlowFloatingPoint)
   public static func ++ (lhs: Tensor, rhs: Tensor) -> Tensor {
     return lhs.concatenated(with: rhs)
   }

Sources/TensorFlow/Operators/Dataset.swift

@@ -41,9 +41,13 @@ func _tensorSeeds(_ seed: Tensor<Int64>) -> (Tensor<Int64>, Tensor<Int64>) {
 /// Represents a potentially large set of elements.
 ///
 /// A `Dataset` can be used to represent an input pipeline as a collection of element tensors.
-@available(*, deprecated, message: """
+@available(
+  *, deprecated,
+  message:
+    """
   Datasets will be removed in S4TF v0.10. Please use the new Batches API instead.
-  """)
+  """
+)
 @frozen
 public struct Dataset<Element: TensorGroup> {
   public let _handle: VariantHandle

Sources/TensorFlow/Operators/LinearAlgebra.swift

@@ -30,7 +30,7 @@ extension Tensor where Scalar: TensorFlowNumeric {
   /// // [1, 2, 3, 4]
   /// ```
   @inlinable
-  @differentiable( where Scalar: TensorFlowFloatingPoint)
+  @differentiable(where Scalar: TensorFlowFloatingPoint)
   public func diagonalPart() -> Tensor {
     precondition(rank >= 2, "The tensor must have at least rank 2.")
     return _Raw.matrixDiagPart(self)
@@ -51,7 +51,7 @@ extension Tensor where Scalar: TensorFlowNumeric {
   /// //  [0, 0, 0, 4]]
   /// ```
   @inlinable
-  @differentiable( where Scalar: TensorFlowFloatingPoint)
+  @differentiable(where Scalar: TensorFlowFloatingPoint)
   public func diagonal() -> Tensor {
     _Raw.matrixDiag(diagonal: self)
   }
@@ -103,7 +103,7 @@ extension Tensor where Scalar: TensorFlowNumeric {
   ///   - superdiagonalCount: The number of superdiagonals to keep. If negative, keep entire upper
   ///     triangle.
   @inlinable
-  @differentiable( where Scalar: TensorFlowFloatingPoint)
+  @differentiable(where Scalar: TensorFlowFloatingPoint)
   public func bandPart(subdiagonalCount: Int, superdiagonalCount: Int) -> Tensor {
     precondition(rank >= 2, "The tensor must have at least rank 2.")
     let lower = Tensor<Int32>(Int32(subdiagonalCount), on: self.device)