我想从此输出中获得三个目标（类的名称）

Question

我对模型进行了培训，然后我预测了句子的结果，我从下面得到了结果，所以我想从这个结果中获得三个课程。 代码：

model = None
with self.strategy.scope():
     # NUMBER OF CLASSES
     model = self.build_model(18)
     model.summary()
    model.load_weights("./saved_weights.h5")
    tokenizer = RobertaTokenizer.from_pretrained(self.MODEL_NAME)
    sentence = "entrepreneurs do you know these 4 golden rules for choosing your partner"
    feat_sent = self.roberta_encode(sentence, tokenizer)
    print(model.predict(feat_sent))

输出：

[[5.9324593e-05 1.3429151e-06 1.8986340e-03 ... 1.3939711e-05
  7.6713404e-03 3.3101414e-06]
 [6.2252270e-06 5.5910685e-05 5.3245033e-04 ... 7.4192496e-05
  6.0618650e-02 1.8048033e-06]
 [1.5471487e-06 6.4954706e-06 8.8266870e-03 ... 1.1088764e-03
  3.0796309e-03 1.4289287e-04]
 ...
 [6.2252229e-06 5.5910707e-05 5.3244922e-04 ... 7.4192525e-05
  6.0618687e-02 1.8048074e-06]
 [5.9324593e-05 1.3429151e-06 1.8986340e-03 ... 1.3939683e-05
  7.6713329e-03 3.3101351e-06]
 [1.4472993e-05 1.6191008e-06 1.4653641e-03 ... 3.5171520e-06
  2.7216172e-03 1.5248560e-08]]

Answer 1

好吧，您可以简单地这样做，但是班级名称是您决定的

results = model.predict(inputs)

# Get max results
preds = results.argmax(axis=1)
label_map = {0: "class_a", 1: "class_b", 2: "class_c"}

# Map array of results via for loop
labels = [label_map[pred] for pred in preds]