测试只保留大数据推送，服务器内存使用情况

algorithm/src/main/java/org/algorithm/core/neural/TensorflowModel.java

@@ -107,28 +107,33 @@ public class TensorflowModel {
                 FloatBuffer.wrap(inputValues)
         );
 
-        float[][] result = null;
+        inputTensor.close();
 
-        // 序列数据
-        if (this.withSequenceInputs){
-            Map<String, Tensor<Integer>> sequenceTensorMap = this.wrapSequenceInputs(sequenceValues, numExamples);
-            this.session.runner();
-
-            result = this.session.runner().feed(this.X, inputTensor)
-                    .feed(this.Char_ids, sequenceTensorMap.get(this.Char_ids))
-                    .feed(this.Pos_ids, sequenceTensorMap.get(this.Pos_ids))
-                    .feed("keep_prob", Tensor.create(1.0f, Float.class))  // dropout保留率
-                    .fetch(this.SOFT_MAX).run().get(0)
-                    .copyTo(new float[numExamples][this.NUM_LABEL]);
-        }else{
-            result = this.session.runner().feed(this.X, inputTensor)
-                    .feed("keep_prob", Tensor.create(1.0f, Float.class))  // dropout保留率
-                    .fetch(this.SOFT_MAX).run().get(0)
-                    .copyTo(new float[numExamples][this.NUM_LABEL]);
+        float[][] fl = new float[numExamples][NUM_LABEL];
+        for (int i = 0; i < numExamples; i++) {
+            for (int j = 0; j < NUM_LABEL; j++) {
+                fl[i][j] = 0.1f;
+            }
         }
-
-        inputTensor.close();
-        return result;
+        return fl;
+//
+//        // 序列数据
+//        if (this.withSequenceInputs){
+//            Map<String, Tensor<Integer>> sequenceTensorMap = this.wrapSequenceInputs(sequenceValues, numExamples);
+//            this.session.runner();
+//
+//            return this.session.runner().feed(this.X, inputTensor)
+//                    .feed(this.Char_ids, sequenceTensorMap.get(this.Char_ids))
+//                    .feed(this.Pos_ids, sequenceTensorMap.get(this.Pos_ids))
+//                    .feed("keep_prob", Tensor.create(1.0f, Float.class))  // dropout保留率
+//                    .fetch(this.SOFT_MAX).run().get(0)
+//                    .copyTo(new float[numExamples][this.NUM_LABEL]);
+//        }else{
+//            return this.session.runner().feed(this.X, inputTensor)
+//                    .feed("keep_prob", Tensor.create(1.0f, Float.class))  // dropout保留率
+//                    .fetch(this.SOFT_MAX).run().get(0)
+//                    .copyTo(new float[numExamples][this.NUM_LABEL]);
+//        }
     }