了解 Keras 预测

Question

我有以下代码：

import tensorflow as tf
from tensorflow import keras
from tensorflow.keras.preprocessing.text import Tokenizer
from tensorflow.keras.preprocessing.sequence import pad_sequences
from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import Dense
from tensorflow.keras.callbacks import EarlyStopping
import numpy as np
from numpy.random import seed
from tensorflow import random

seed(42)
random.set_seed(43)

X = [
    'may it all be fine in the world',
    'this is not for me',
    'pffff ugly bike',
    'dropping by to say leave me alone',
    'getting sarcastic by now'
    'how would one satisfy his or her needs when the earth is boiling'
]

y = [1,2,4,5,3]

tokenizer = Tokenizer(num_words = 13)
tokenizer.fit_on_texts(X)
X_train_seq = tokenizer.texts_to_sequences(X)


X_train_seq_padded = pad_sequences(X_train_seq, maxlen = 15)

model = Sequential()
model.add(Dense(16, input_dim = 15, activation = 'relu', name = 'hidden-1'))
model.add(Dense(16, activation = 'relu', name = 'hidden-2'))
model.add(Dense(16, activation = 'relu', name = 'hidden-3'))
model.add(Dense(5, activation='softmax', name = 'output_layer'))

model.compile(optimizer = 'adam', loss = 'categorical_crossentropy', metrics=['accuracy'])

class CustomCallback(keras.callbacks.Callback):
    def on_epoch_end(self, epoch, logs=None):
        print('finished an epoch')
        zin = 'dropping by to say leave her alone'
        zin = tokenizer.texts_to_sequences(zin)
        zin = pad_sequences(zin, maxlen = 15)
        print(model.predict(zin))
        print(np.argmax(model.predict(zin), axis=-1))
callbacks = [EarlyStopping(monitor = 'accuracy', patience = 5, mode = 'max'), CustomCallback()]

from sklearn.preprocessing import LabelBinarizer
encoder = LabelBinarizer()
y = encoder.fit_transform(y)

history = model.fit(X_train_seq_padded, y, epochs = 100, batch_size = 100, callbacks = callbacks)

我希望在回调 model.predict() 内部会产生类似的结果（因为有 5 个可能的类）：

[0.4534534, 0.5634246, 0.0045623, 0.0004536, 0.0000056]

和 np.argmax(model.predict(zin), axis=-1) 在单个数字 1、2、3、4 或 5 中。

但是我收到的输出（显示一个时期）是：

我必须如何解释这一点，如何过滤掉模型预测句子所属的实际类别？

Answer 1

print(model.predict(zin)[0])
print(np.argmax(model.predict(zin)[0], axis=-1))

这应该给你正确的价值。

tf 模型设计为批量工作，而不是作为单一用途，因此它为您提供输出列表，但由于您的输入是单个项目，它只是将该项目通过 NN 推送 n 次，因此是相同输出列表