java日常學習四：調用機器學習分類模型

// 下載模型
private static final SavedModelBundle modelBundle = SavedModelBundle.load(模型路徑,"serve");
// session
private static final Session tfSession = modelBundle.session();
// 預測
private static final Operation operationPredict = modelBundle.graph().operation("score/my_predict");
// 輸出
private static final Output output = new Output(operationPredict,0);
// Tensor
private static final Tensor keep_prob = Tensor.create(Float.parseFloat("0.6"));

// 輸入轉換成向量，a爲向量

Tensor input_x = Tensor.create(a);
// 預測輸出，類目對應編號
Tensor out = tfSession.runner().feed("input_x", input_x).feed("keep_prob",keep_prob).fetch(output).run().get(0);
long[] temp = new long[1];
out.copyTo(temp);
short reskey = (short) temp[0];

調用線上模型：

// 輸入長度---並轉化成向量
List<Object> requestData = new ArrayList<>(Collections.nCopies(輸入向量長度,0));

// 預測結果

Object responseData = iWpaiDLPredictOnlineService.tensorflowServingPredictOnline(線上服務編號,requestData);