我的代码：
import numpy as np
from keras import Input, Model
from keras.layers import LSTM, Dense
input_texts = []
target_texts = []
input_characters = set()
target_characters = set()
with open('catalan.txt', 'r', encoding = 'utf-8') as f:
lines = f.read().split('n')
for line in lines[: min(653, len(lines) - 1)]:
input_text, target_text = line.split('t')
target_text = 't' + target_text + 'n'
input_texts.append(input_text)
target_texts.append(target_text)
for char in input_text:
if char not in input_characters:
input_characters.add(char)
for char in target_text:
if char not in target_characters:
target_characters.add(char)
input_characters = sorted(list(input_characters))
target_characters = sorted(list(target_characters))
num_encoder_tokens = len(input_characters)
num_decoder_tokens = len(target_characters)
max_encoder_seq_length = max([len(txt) for txt in input_texts])
max_decoder_seq_length = max([len(txt) for txt in target_texts])
input_token_index = dict(
[(char, i) for i, char in enumerate(input_characters)])
target_token_index = dict(
[(char, i) for i, char in enumerate(target_characters)])
encoder_input_data = np.zeros(
(len(input_texts), max_encoder_seq_length, num_encoder_tokens),
dtype = 'float32')
decoder_input_data = np.zeros(
(len(input_texts), max_decoder_seq_length, num_decoder_tokens),
dtype = 'float32')
decoder_target_data = np.zeros(
(len(input_texts), max_decoder_seq_length, num_decoder_tokens),
dtype = 'float32')
for i, (input_text, target_text) in enumerate(zip(input_texts, target_texts)):
for t, char in enumerate(input_text):
encoder_input_data[i, t, input_token_index[char]] = 1.
for t, char in enumerate(target_text):
decoder_input_data[i, t, target_token_index[char]] = 1.
if t > 0:
decoder_target_data[i, t - 1, target_token_index[char]] = 1.
latent_dim = 10
batch_size = 256
epochs = 10
encoder_inputs = Input(shape = (None, num_encoder_tokens))
encoder = LSTM(latent_dim, return_state = True)
encoder_outputs, state_h, state_c = encoder(encoder_inputs)
encoder_states = [state_h, state_c]
decoder_inputs = Input(shape = (None, num_decoder_tokens))
decoder_lstm = LSTM(latent_dim, return_sequences = True, return_state = True)
decoder_outputs, _, _ = decoder_lstm(decoder_inputs,
initial_state = encoder_states)
decoder_dense = Dense(num_decoder_tokens, activation = 'softmax')
decoder_outputs = decoder_dense(decoder_outputs)
model = Model([encoder_inputs, decoder_inputs], decoder_outputs)
model.compile(optimizer = 'rmsprop', loss = 'categorical_crossentropy')
model.fit([encoder_input_data, decoder_input_data], decoder_target_data,
batch_size = batch_size, epochs = epochs, validation_split = 0.2)
encoder_model = Model(encoder_inputs, encoder_states)
decoder_state_input_h = Input(shape = (latent_dim,))
decoder_state_input_c = Input(shape = (latent_dim,))
decoder_states_inputs = [decoder_state_input_h, decoder_state_input_c]
decoder_outputs, state_h, state_c = decoder_lstm(
decoder_inputs, initial_state = decoder_states_inputs)
decoder_states = [state_h, state_c]
decoder_outputs = decoder_dense(decoder_outputs)
decoder_model = Model(
[decoder_inputs] +
decoder_states_inputs,
[decoder_outputs] +
decoder_states)
# Reverse-lookup token index to decode sequences back to
# something readable.
reverse_input_char_index = dict(
(i, char) for char, i in input_token_index.items())
reverse_target_char_index = dict(
(i, char) for char, i in target_token_index.items())
def decode_sequence(input_seq):
states_value = encoder_model.predict(input_seq)
target_seq = np.zeros((1, 1, num_decoder_tokens))
target_seq[0, 0, target_token_index['t']] = 1.
stop_condition = False
decoded_sentence = ''
while not stop_condition:
output_tokens, h, c = decoder_model.predict(
[target_seq] + states_value)
# Sample a token
sampled_token_index = np.argmax(output_tokens[0, -1, :])
sampled_char = reverse_target_char_index[sampled_token_index]
decoded_sentence += sampled_char
if (sampled_char == 'n' or
len(decoded_sentence) > max_decoder_seq_length):
stop_condition = True
target_seq = np.zeros((1, 1, num_decoder_tokens))
target_seq[0, 0, sampled_token_index] = 1.
# Update states
states_value = [h, c]
return decoded_sentence
for seq_index in range(5):
input_seq = encoder_input_data[seq_index: seq_index + 1]
decoded_sentence = decode_sequence(input_seq)
print('n')
print('Input sentence:', input_texts[seq_index])
print('Decoded sentence:', decoded_sentence)
输出：
Using TensorFlow backend.
2020-03-06 16:37:17.569143: I tensorflow/compiler/xla/service/service.cc:168] XLA service 0x7fc6781e2ee0 initialized for platform Host (this does not guarantee that XLA will be used). Devices:
2020-03-06 16:37:17.569165: I tensorflow/compiler/xla/service/service.cc:176] StreamExecutor device (0): Host, Default Version
Train on 521 samples, validate on 131 samples
Epoch 1/10
256/521 [=============>................] - ETA: 1s - loss: 1.3404
512/521 [============================>.] - ETA: 0s - loss: 1.3235
521/521 [==============================] - 2s 4ms/step - loss: 1.3269 - val_loss: 2.2806
Epoch 2/10
256/521 [=============>................] - ETA: 0s - loss: 1.3238
512/521 [============================>.] - ETA: 0s - loss: 1.3232
521/521 [==============================] - 1s 1ms/step - loss: 1.3226 - val_loss: 2.2743
Epoch 3/10
256/521 [=============>................] - ETA: 0s - loss: 1.3432
512/521 [============================>.] - ETA: 0s - loss: 1.3204
521/521 [==============================] - 1s 1ms/step - loss: 1.3192 - val_loss: 2.2671
Epoch 4/10
256/521 [=============>................] - ETA: 0s - loss: 1.3363
512/521 [============================>.] - ETA: 0s - loss: 1.3180
521/521 [==============================] - 1s 1ms/step - loss: 1.3153 - val_loss: 2.2586
Epoch 5/10
256/521 [=============>................] - ETA: 0s - loss: 1.2933
512/521 [============================>.] - ETA: 0s - loss: 1.3102
521/521 [==============================] - 1s 1ms/step - loss: 1.3105 - val_loss: 2.2467
Epoch 6/10
256/521 [=============>................] - ETA: 0s - loss: 1.3062
512/521 [============================>.] - ETA: 0s - loss: 1.3085
521/521 [==============================] - 1s 2ms/step - loss: 1.3038 - val_loss: 2.2313
Epoch 7/10
256/521 [=============>................] - ETA: 0s - loss: 1.3044
512/521 [============================>.] - ETA: 0s - loss: 1.2919
521/521 [==============================] - 1s 1ms/step - loss: 1.2947 - val_loss: 2.2081
Epoch 8/10
256/521 [=============>................] - ETA: 0s - loss: 1.2874
512/521 [============================>.] - ETA: 0s - loss: 1.2801
521/521 [==============================] - 1s 1ms/step - loss: 1.2816 - val_loss: 2.1818
Epoch 9/10
256/521 [=============>................] - ETA: 0s - loss: 1.2862
512/521 [============================>.] - ETA: 0s - loss: 1.2708
521/521 [==============================] - 1s 1ms/step - loss: 1.2670 - val_loss: 2.1564
Epoch 10/10
256/521 [=============>................] - ETA: 0s - loss: 1.2387
512/521 [============================>.] - ETA: 0s - loss: 1.2506
521/521 [==============================] - 1s 1ms/step - loss: 1.2528 - val_loss: 2.1281
Input sentence: Wow!
Decoded sentence: t
Input sentence: Really?
Decoded sentence: t
Input sentence: Thanks.
Decoded sentence: t
Input sentence: Goodbye!
Decoded sentence: t
Input sentence: Hurry up.
Decoded sentence: t
catalan.txt包含以下结构的文本：
Wow! Carai!
Really? De veritat?
Thanks. Gràcies!
Goodbye! Adéu!
Hurry up. Afanya't.
Too late. Massa tard.
为什么我总是得到t？我想这一定是一个英文句子的翻译。它怎么了？
正如@Recessionive在评论中回答的那样：增加时代。
我用1000进行了测试，它有效，没有改变其他参数。
此外：通过调整这些其他参数，结果会随着时间的减少而变得更好。
这意味着：在纠正了while中的return之后，代码似乎是正确的，如@h4z3所示。