Adding Matrix Tensor with auto shape / auto flat functionality.

GUIDE.md

@@ -178,6 +178,16 @@ let datastr = tensor.strings
 let str = datastr.map { $0.string }
 ```
 
+### Matrix
+
+Since Perfect-TensorFlow v1.2.1, however, you can apply a multi-dimensional array to a tensor as a matrix without considering the shape or dimensions. The equivalent example of above will look like:
+
+``` swift
+let M = try TF.Tensor.Matrix([[1, 2], [3, 4]])
+```
+
+⚠️**NOTES**⚠️ Element in a Matrix must be number!
+
 ### Raw Data of a Tensor
 
 To access raw data of a tensor, you can use either `data` property, or `withDataPointer()` method with better performance but tricky pointer operations - and property of `bytesCount` and `type` will be useful in this case:

GUIDE.zh_CN.md

@@ -176,6 +176,16 @@ let datastr = tensor.strings
 let str = datastr.map { $0.string }
 ```
 
+### Matrix
+
+自从 Perfect-TensorFlow v1.2.1 版本开始，您可以直接输入矩阵作为张量了！上述例子现在可以改写为：
+
+``` swift
+let M = try TF.Tensor.Matrix([[1, 2], [3, 4]])
+```
+
+⚠️**NOTES**⚠️ Element in a Matrix must be number!
+
 ### 访问张量原始数据
 
 如果您希望在程序中访问张量的原始数据，您可以使用`data`属性，或者调用`withDataPointer()`方法通过指针进行访问（虽然有些技巧性难度，但是性能要快很多）——这种情况下`byteCount`属性（张量数据在内存中的大小）和`type`（张量数据类型）会有很大用处：

README.md

@@ -161,14 +161,8 @@ Here is an other simple example of matrix operations in Perfect TensorFlow:
 | 3 4 | * |0 0| = |0 3|
 */
 // input the matrix.
-// *NOTE* no matter how many dimensions a matrix may have,
-// the matrix should always input as an flattened array
-let srcA:[Float] = [[1, 2], [3, 4]].flatMap { $0 }
-let srcB:[Float] = [[0, 0], [1, 0]].flatMap { $0 }
-
-// create tensors for these matrices
-let tA = try TF.Tensor.Array(dimensions: [2,2], value: srcA)
-let tB = try TF.Tensor.Array(dimensions: [2,2], value: srcB)
+let tA = try TF.Tensor.Matrix([[1, 2], [3, 4]])
+let tB = try TF.Tensor.Matrix([[0, 0], [1, 0]])
 
 // adding tensors to graph
 let g = try TF.Graph()

README.zh_CN.md

@@ -160,13 +160,8 @@ print(s2)
 | 3 4 | * |0 0| = |0 3|
 */
 // 输入矩阵
-// *注意* 不管目标矩阵的维度到底是多少，输入时必须按照扁平化处理，即所有元素一个挨一个列出
-let srcA:[Float] = [[1, 2], [3, 4]].flatMap { $0 }
-let srcB:[Float] = [[0, 0], [1, 0]].flatMap { $0 }
-
-// 根据矩阵创建张量
-let tA = try TF.Tensor.Array(dimensions: [2,2], value: srcA)
-let tB = try TF.Tensor.Array(dimensions: [2,2], value: srcB)
+let tA = try TF.Tensor.Matrix([[1, 2], [3, 4]])
+let tB = try TF.Tensor.Matrix([[0, 0], [1, 0]])
 
 // 将张量转化为流程图节点
 let g = try TF.Graph()

Sources/PerfectTensorFlow/PerfectTensorFlow.swift

@@ -34,6 +34,36 @@ public extension Array {
   public func flat() -> Array<Any> {
     return Array.Flat(self)
   }
+
+  public var shape: [Int] {
+    var _shape = [Int]()
+    var a = self as Array<Any>
+    while a.count > 0 {
+      _shape.append(a.count)
+      if let b = a as? Array<Array<Any>> , let c = b.first {
+        a = c
+      } else {
+        break
+      }//end if
+    }//end while
+    return _shape
+  }//end var
+
+  public func column(index: Int) -> Array<Any> {
+    var b = [Any]()
+    let s = shape
+    guard s.count > 1, index > -1, index < s[1],
+      let a = self as? Array<Array<Any>> else {
+        if index > -1 && index < self.count {
+          b.append(self[index])
+        }//end if
+        return b
+    }//end guard
+    a.forEach { c in
+      b.append(c[index])
+    }//next
+    return b
+  }//end func
 }
 
 typealias SwiftArray<T> = Array<T>
@@ -568,6 +598,41 @@ public class TensorFlow {
       }//end if
     }//end tensor
 
+    /// Create a tensor from a Matrix. *NOTE* Element must be number
+    /// - parameters:
+    ///   - matrix: Matrix in form of Array[Array[Array ... Array[Number]]]
+    ///   - throws: Panic.FAULT
+    /// - returns: tensor
+    public static func Matrix(_ matrix: Array<Any>) throws -> Tensor {
+      let shape = matrix.shape.map { Int64($0) }
+      let flattened = matrix.flat()
+      let f: [Float]
+      if flattened is [UInt8], let i = flattened as? [UInt8] {
+        f = i.map { Float($0) }
+      } else if flattened is [UInt32], let i = flattened as? [UInt32] {
+        f = i.map { Float($0) }
+      } else if flattened is [UInt], let i = flattened as? [UInt] {
+        f = i.map { Float($0) }
+      } else if flattened is [UInt64], let i = flattened as? [UInt64] {
+        f = i.map { Float($0) }
+      } else if flattened is [Int8], let i = flattened as? [Int8] {
+        f = i.map { Float($0) }
+      } else if flattened is [Int32], let i = flattened as? [Int32] {
+        f = i.map { Float($0) }
+      } else if flattened is [Int], let i = flattened as? [Int] {
+        f = i.map { Float($0) }
+      } else if flattened is [Int64], let i = flattened as? [Int64] {
+        f = i.map { Float($0) }
+      } else if flattened is [Float], let i = flattened as? [Float] {
+        f = i
+      } else if flattened is [Double], let i = flattened as? [Double] {
+        f = i.map { Float($0) }
+      } else {
+        throw Panic.FAULT(reason: "Matrix Element Must Be Number")
+      }//end if
+      return try Array(dimensions: shape, value: f)
+    }
+
     /// Create a tensor by passing its value
     /// type can be Int, Float,... or String / Data
     /// - parameters:
@@ -577,6 +642,7 @@ public class TensorFlow {
     ///   Tensor
     /// - throws:
     ///   Panic
+
     public static func Array<T> (dimensions:[Int64], value: [T]) throws -> Tensor {
       let count = Int(dimensions.reduce(1) { $0 * $1 })
       guard let _ = TFLib.libDLL,

Tests/PerfectTensorFlowTests/PerfectTensorFlowTests.swift

@@ -144,9 +144,20 @@ class PerfectTensorFlowTests: XCTestCase {
     ("testBasicExpress", testBasicExpress),
     ("testLabels", testLabels),
     ("testSessionLeak", testSessionLeak),
-    ("testGradients", testGradients)
+    ("testGradients", testGradients),
+    ("testMatrix", testMatrix),
+    ("testBasicImproved",testBasicImproved)
   ]
 
+  func testMatrix() {
+    let x = [[1, 2, 3], [4, 5, 6]]
+    let y = [[[1,2],[3,4],[5,6]],[[7,8],[9,10],[11,12]],[[13,14],[15,16],[17,18]],[[19,20],[21,22],[23,24]]]
+    XCTAssertEqual(x.shape, [2, 3])
+    XCTAssertEqual(y.shape, [4, 3, 2])
+    XCTAssertEqual(x[1][2], 6)
+    XCTAssertEqual(x.column(index: 1) as! [Int], [2, 5])
+  }
+
   func testGradients() {
     do {
       let grad = TestGradients()
@@ -353,6 +364,29 @@ class PerfectTensorFlowTests: XCTestCase {
 
   }
 
+  func testBasicImproved() {
+    do {
+      /*
+       Matrix Test:
+       | 1 2 |  |0 1|  |0 1|
+       |     |* |   |= |   |
+       | 3 4 |  |0 0|  |0 3|
+       */
+      let tA = try TF.Tensor.Matrix([[1, 2], [3, 4]])
+      let tB = try TF.Tensor.Matrix([[0, 0], [1, 0]])
+      let g = try TF.Graph()
+      let A = try g.const(tensor: tA, name: "Const_0")
+      let B = try g.const(tensor: tB, name: "Const_1")
+      let v = try g.matMul(l: A, r: B, name: "v", transposeB: true)
+      let o = try g.runner().fetch(v).addTarget(v).run()
+      let m:[Float] = try o[0].asArray()
+      let r:[Float] = [0, 1, 0, 3]
+      XCTAssertEqual(m, r)
+    }catch {
+      XCTFail("improved: \(error)")
+    }
+  }
+
   func testBasicExpress() {
     do {
       /*