darcwader

from tensorflow.python.client import device_lib
print(device_lib.list_local_devices())

darcwader

#preference is install for scrip runfile local. dont use apt
sudo apt purge nvidia-396 nvidia-*
sudo /etc/init.d/sddm stop
sudo /etc/init.d/gdm stop
sudo killall Xorg
sudo apt purge xserver-xorg-video-nouveau*
sudo vim /etc/modprobe.d/blacklist-nouveau.conf
#blacklist nouveau
#options nouveau modeset=0
sudo sh ./cuda_9.2.148_396.37_linux.run

darcwader

//make sure to have textField outlet
let vector = spam.tfidf(sentence: self.textField.text ?? "")
let mlarray = spam.multiarray(vector: vector)
let res =  try model.prediction(message: mlarray)
print(res.spam_or_not)

print(res.classProbability)

darcwader

func multiarray(vector:[Int:Double]) -> MLMultiArray {
    let array = try! MLMultiArray(shape: [NSNumber(integerLiteral: self.vocabulary.count)], dataType: .double)
    for (key, value) in vector {
        array[key] = NSNumber(floatLiteral: value)
    }
    return array
}

darcwader

func idf(word:String) -> Double {
    if let pos = self.vocabulary[word] {
        return self.idf[pos]
    } else {
        return Double(0.0)
    }
}


func tfidf(sentence:String) -> [Int:Double] {

darcwader

func countVector(sentence:String) -> [Int:Int]? {
    var vec = [Int:Int]()
    for word in self.tokenize(sentence) {
        if let pos = self.vocabulary[word] {
            if let i = vec[pos] {
                vec[pos] = i+1
            } else {
                vec[pos] = 1
            }
        }

darcwader

func tokenize(_ message:String) -> [String] {
        let trimmed = message.lowercased().trimmingCharacters(in: CharacterSet(charactersIn: "!\"#$%&'()*+,-./:;<=>?@[\\]^_`{|}~"))
        let tokens = trimmed.components(separatedBy: CharacterSet.whitespaces)
        return tokens
}

darcwader

init() {
        let wordsPath = Bundle.main.url(forResource:"words_array", withExtension:"json")
        do {
            let wordsData = try Data(contentsOf: wordsPath!)
            if let wordsDict = try JSONSerialization.jsonObject(with: wordsData, options: []) as? [String:Int] {
                self.vocabulary = wordsDict
            }
        } catch {
            fatalError("oops could not load words_array")
        }

darcwader

class Spam {
    var idf = [Double]()
    var vocabulary = [String:Int]()
    var norm:Bool = true
}

darcwader

import coremltools
coreml_model = coremltools.converters.sklearn.convert(spam_detector, "message", "spam_or_not")

#set parameters of the model
coreml_model.short_description = "Classify whether message is spam or not"
coreml_model.input_description["message"] = "TFIDF of message to be classified"
coreml_model.output_description["spam_or_not"] = "Whether message is spam or not"
#save the model
coreml_model.save("SpamMessageClassifier.mlmodel")