TensorFlow聊天模型训练报错：Sequential层输入张量数量不符

错误信息 #

训练时触发如下错误：

ValueError: Layer "sequential" expects 1 input(s), but it received 37 input tensors. Inputs received: [<tf.Tensor 'IteratorGetNext:0' shape=(None,) dtype=int32>, ...]

问题代码与配置 #

训练代码 #

import random
import json
import pickle
import numpy as np

import nltk
from nltk.stem import WordNetLemmatizer

# nltk.download('punkt')
# nltk.download('wordnet')

import tensorflow as tf
from tensorflow import *
import keras
from keras import layers

import os

os.environ['TF_CPP_MIN_LOG_LEVEL'] = '1'  # 屏蔽CPU类型导致的版本错误

lemmatizer = WordNetLemmatizer()

intents = json.loads(open('intents.json').read())

words = []
classes = []
documents = []
ignore_letters = ['?', '!', '.', ',']

for intent in intents['intents']:
    for pattern in intent['patterns']:
        word_list = nltk.word_tokenize(pattern)
        words.extend(word_list)
        documents.append((word_list, intent['tag']))
        if intent['tag'] not in classes:
            classes.append(intent['tag'])

words = [lemmatizer.lemmatize(word) for word in words if word not in ignore_letters]
words = sorted(set(words))

classes = sorted(set(classes))

pickle.dump(words, open('words.pkl', 'wb'))
pickle.dump(words, open('classes.pkl', 'wb'))  # 此处存在错误，应为classes

training = []
output_empty = [0] * len(classes)

for document in documents:
    bag = []
    word_patterns = document[0]
    word_patterns = [lemmatizer.lemmatize(word.lower()) for word in word_patterns]
    for word in words:
        bag.append(1) if word in word_patterns else bag.append(0)

    output_row = list(output_empty)
    output_row[classes.index(document[1])] = 1
    training.append([bag, output_row])

data_tensor = tf.ragged.constant(training)  
data_tensor = data_tensor.to_tensor()

random.shuffle(data_tensor)

data_tensor = np.array(data_tensor)

train_x = list(data_tensor[:, 0])
train_y = list(data_tensor[:, 1])

# train_x = np.array(train_x, dtype=int)
# train_y = np.array(train_y, dtype=int)

model = tf.keras.Sequential()

model.add(layers.Dense(128, input_shape=(len(train_x[0]),), activation='relu'))
model.add(layers.Dropout(0.5))
model.add(layers.Dense(64, activation='relu'))
model.add(layers.Dropout(0.5))
model.add(layers.Dense(len(train_y[0]), activation='softmax'))

sgd = tf.keras.optimizers.experimental.SGD(learning_rate=0.01, momentum=0.0, nesterov=True, weight_decay=None, global_clipnorm=None, ema_momentum=0.99, ema_overwrite_frequency=None, name='SGD',)

model.compile(loss='categorical_crossentropy', optimizer=sgd, metrics=['accurancy'])  # 拼写错误：accurancy -> accuracy

model.fit(train_x, train_y, batch_size=5, epochs=10, verbose=1)
model.save('chatbot_model.model')

print("\nDone\n")

model.load('chatbot_model.model')  # 方法错误，应为load_model

intents.json结构示例 #

{"intents":  [
{"tag": "greetings",
"patterns": ["hello", "hey", "hi", "good day", "Greetings", "what's up?", "how is it going?"],
"responses": ["Hello!", "Hey!", "What can I do for you?"]
},

{"tag": "goodbye",
"patterns": ["cya", "See you later", "Goodbye", "I am Leaving", "Have a Good day", "bye", "cao", "see ya"],
"responses": ["Sad to see you go :(", "Talk to you later", "Goodbye!"]
},

{"tag": "name",
"patterns": ["what is your name", "what should i call you", "what should i call you", "what your name?", "who are you?", "who are you?", "can you tell me your name"],
"responses": ["Hello!", "Hey!", "What can I do for you?"]
}
]}

错误原因与修复步骤 #

核心错误原因 #

train_x和train_y被转为列表形式的一维数组，TensorFlow会将列表中的每个元素视为独立输入，导致模型收到多个输入张量，与第一层定义的单输入形状不匹配。

具体修复点 #

1. 转换输入数据为二维numpy数组
   取消注释并修正数据类型转换代码，确保输入是模型预期的二维张量：

train_x = np.array(train_x, dtype=np.float32)
train_y = np.array(train_y, dtype=np.float32)

2. 修正指标拼写错误
   将model.compile中的'accurancy'改为正确的'accuracy'：

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

3. 修正模型加载方法
   model.load不是合法方法，替换为TensorFlow官方加载函数：

model = tf.keras.models.load_model('chatbot_model.model')

4. 修正pickle保存错误
   保存classes时误存了words，修正为：

pickle.dump(classes, open('classes.pkl', 'wb'))

5. 修正循环缩进问题
   原代码中for intent和for pattern的内部代码没有缩进，这会导致语法错误，需按Python缩进规范调整（已在上面的代码示例中修正）。

修复后训练流程 #

完成上述修改后，模型将正确识别输入形状，可正常执行训练流程。

内容的提问来源于stack exchange，提问作者EinShadow