Changing default download location for BostonHousing and CIFAR-10 datasets (#430)

* Replace links BostonHousing and CIFAR-10 datasets, along with RoBERTa checkpoint.

* Formatting BostonHousing.swift.

Author: BradLarson

Datasets/BostonHousing/BostonHousing.swift

@@ -22,7 +22,7 @@ import ModelSupport
 import TensorFlow
 
 public struct BostonHousing {
-    public let trainPercentage:Float = 0.8
+    public let trainPercentage: Float = 0.8
     public let numRecords: Int
     public let numColumns: Int
     public let numTrainRecords: Int
@@ -33,35 +33,41 @@ public struct BostonHousing {
     public let yTest: Tensor<Float>
 
     static func downloadBostonHousingIfNotPresent() -> String {
-        let remoteURL = URL(string: "https://archive.ics.uci.edu/ml/machine-learning-databases/housing/")!
-        let localURL = DatasetUtilities.defaultDirectory.appendingPathComponent("BostonHousing", isDirectory: true)
+        let remoteURL = URL(
+            string: "https://storage.googleapis.com/s4tf-hosted-binaries/datasets/BostonHousing/")!
+        let localURL = DatasetUtilities.defaultDirectory.appendingPathComponent(
+            "BostonHousing", isDirectory: true)
 
         let downloadPath = localURL.path
         let directoryExists = FileManager.default.fileExists(atPath: downloadPath)
         let contentsOfDir = try? FileManager.default.contentsOfDirectory(atPath: downloadPath)
         let directoryEmpty = (contentsOfDir == nil) || (contentsOfDir!.isEmpty)
-        
+
         if !directoryExists || directoryEmpty {
             let _ = DatasetUtilities.downloadResource(
                 filename: "housing", fileExtension: "data",
                 remoteRoot: remoteURL, localStorageDirectory: localURL,
                 extract: false)
         }
 
-        return try! String(contentsOf: localURL.appendingPathComponent("housing.data"), encoding: String.Encoding.utf8)
+        return try! String(
+            contentsOf: localURL.appendingPathComponent("housing.data"),
+            encoding: String.Encoding.utf8)
     }
-    
+
     public init() {
         let data = BostonHousing.downloadBostonHousingIfNotPresent()
 
         // Convert Space Separated CSV with no Header
-        let dataRecords: [[Float]] = data.split(separator: "\n").map{ String($0).split(separator: " ").compactMap{ Float(String($0)) } }
+        let dataRecords: [[Float]] = data.split(separator: "\n").map {
+            String($0).split(separator: " ").compactMap { Float(String($0)) }
+        }
 
         let numRecords = dataRecords.count
         let numColumns = dataRecords[0].count
 
-        let dataFeatures = dataRecords.map{ Array($0[0..<numColumns-1]) }
-        let dataLabels = dataRecords.map{ Array($0[(numColumns-1)...]) }
+        let dataFeatures = dataRecords.map { Array($0[0..<numColumns - 1]) }
+        let dataLabels = dataRecords.map { Array($0[(numColumns - 1)...]) }
 
         self.numRecords = numRecords
         self.numColumns = numColumns
@@ -73,15 +79,17 @@ public struct BostonHousing {
         let yTrain = Array(Array(dataLabels[0..<numTrainRecords]).joined())
         let yTest = Array(Array(dataLabels[numTrainRecords...]).joined())
 
-        let xTrainDeNorm = Tensor<Float>(xTrain).reshaped(to: TensorShape([numTrainRecords, numColumns-1]))
-        let xTestDeNorm = Tensor<Float>(xTest).reshaped(to: TensorShape([numTestRecords, numColumns-1]))
+        let xTrainDeNorm = Tensor<Float>(xTrain).reshaped(
+            to: TensorShape([numTrainRecords, numColumns - 1]))
+        let xTestDeNorm = Tensor<Float>(xTest).reshaped(
+            to: TensorShape([numTestRecords, numColumns - 1]))
 
         // Normalize
         let mean = xTrainDeNorm.mean(alongAxes: 0)
         let std = xTrainDeNorm.standardDeviation(alongAxes: 0)
 
-        self.xTrain = (xTrainDeNorm - mean)/std
-        self.xTest = (xTestDeNorm - mean)/std
+        self.xTrain = (xTrainDeNorm - mean) / std
+        self.xTest = (xTestDeNorm - mean) / std
         self.yTrain = Tensor<Float>(yTrain).reshaped(to: TensorShape([numTrainRecords, 1]))
         self.yTest = Tensor<Float>(yTest).reshaped(to: TensorShape([numTestRecords, 1]))
     }

Datasets/CIFAR10/CIFAR10.swift

@@ -31,7 +31,7 @@ public struct CIFAR10: ImageClassificationDataset {
         self.init(
             batchSize: batchSize,
             remoteBinaryArchiveLocation: URL(
-                string: "https://www.cs.toronto.edu/~kriz/cifar-10-binary.tar.gz")!, 
+                string: "https://storage.googleapis.com/s4tf-hosted-binaries/datasets/CIFAR10/cifar-10-binary.tar.gz")!, 
             normalizing: true)
     }
 

Models/Text/BERT.swift

@@ -510,7 +510,7 @@ extension BERT {
         /// The URL where this pre-trained model can be downloaded from.
         public var url: URL {
             let bertPrefix = "https://storage.googleapis.com/bert_models/2018_"
-            let robertaPrefix = "https://www.dropbox.com/s"
+            let robertaPrefix = "https://storage.googleapis.com/s4tf-hosted-binaries/checkpoints/Text/RoBERTa"
             let albertPrefix = "https://storage.googleapis.com/tfhub-modules/google/albert"
             switch self {
             case .bertBase(false, false):
@@ -530,9 +530,9 @@ extension BERT {
             case .bertLarge(true, true):
                 return URL(string: "\(bertPrefix)05_30/\(subDirectory).zip")!
             case .robertaBase:
-                return URL(string: "\(robertaPrefix)/12ymhgwbfxm2ozf/base.zip?dl=1")!
+                return URL(string: "\(robertaPrefix)/base.zip")!
             case .robertaLarge:
-                return URL(string: "\(robertaPrefix)/jf6kxmdvxyfl4wz/large.zip?dl=1")!
+                return URL(string: "\(robertaPrefix)/large.zip")!
             case .albertBase, .albertLarge, .albertXLarge, .albertXXLarge:
                 return URL(string: "\(albertPrefix)_\(subDirectory)/1.tar.gz")!
             }