 Lihat di TensorFlow.org |  Jalankan di Google Colab |  Lihat sumber di GitHub |  Unduh buku catatan |
Notebook ini melatih model urutan ke urutan (seq2seq) untuk terjemahan Bahasa Spanyol ke Bahasa Inggris. Ini adalah contoh lanjutan yang mengasumsikan pengetahuan tentang urutan ke model urutan.
Setelah melatih model dalam buku catatan ini, Anda dapat memasukkan kalimat bahasa Spanyol, seperti "¿todavia estan en casa?" , dan kembalikan terjemahan bahasa Inggris: "apakah Anda masih di rumah?"
Kualitas terjemahan layak untuk contoh mainan, tetapi plot perhatian yang dihasilkan mungkin lebih menarik. Ini menunjukkan bagian kalimat input mana yang menarik perhatian model saat menerjemahkan:
import tensorflow as tf
import matplotlib.pyplot as plt
import matplotlib.ticker as ticker
from sklearn.model_selection import train_test_split
import unicodedata
import re
import numpy as np
import os
import io
import time
Unduh dan siapkan dataset
Kami akan menggunakan dataset bahasa yang disediakan oleh http://www.manythings.org/anki/ Dataset ini berisi pasangan terjemahan bahasa dalam format:
May I borrow this book? ¿Puedo tomar prestado este libro?
Ada berbagai bahasa yang tersedia, tetapi kami akan menggunakan dataset bahasa Inggris-Spanyol. Untuk kenyamanan, kami telah meng-host salinan dataset ini di Google Cloud, tetapi Anda juga dapat mengunduh salinan Anda sendiri. Setelah mengunduh dataset, berikut adalah langkah-langkah yang akan kami ambil untuk menyiapkan data:
- Tambahkan token awal dan akhir untuk setiap kalimat.
- Bersihkan kalimat dengan menghapus karakter khusus.
- Buat indeks kata dan membalikkan indeks kata (kamus kamus dari kata → id dan id → kata).
- Pad setiap kalimat dengan panjang maksimal.
# Download the file
path_to_zip = tf.keras.utils.get_file(
'spa-eng.zip', origin='http://storage.googleapis.com/download.tensorflow.org/data/spa-eng.zip',
extract=True)
path_to_file = os.path.dirname(path_to_zip)+"/spa-eng/spa.txt"
Downloading data from http://storage.googleapis.com/download.tensorflow.org/data/spa-eng.zip 2646016/2638744 [==============================] - 0s 0us/step
# Converts the unicode file to ascii
def unicode_to_ascii(s):
return ''.join(c for c in unicodedata.normalize('NFD', s)
if unicodedata.category(c) != 'Mn')
def preprocess_sentence(w):
w = unicode_to_ascii(w.lower().strip())
# creating a space between a word and the punctuation following it
# eg: "he is a boy." => "he is a boy ."
# Reference:- https://stackoverflow.com/questions/3645931/python-padding-punctuation-with-white-spaces-keeping-punctuation
w = re.sub(r"([?.!,¿])", r" \1 ", w)
w = re.sub(r'[" "]+', " ", w)
# replacing everything with space except (a-z, A-Z, ".", "?", "!", ",")
w = re.sub(r"[^a-zA-Z?.!,¿]+", " ", w)
w = w.strip()
# adding a start and an end token to the sentence
# so that the model know when to start and stop predicting.
w = '<start> ' + w + ' <end>'
return w
en_sentence = u"May I borrow this book?"
sp_sentence = u"¿Puedo tomar prestado este libro?"
print(preprocess_sentence(en_sentence))
print(preprocess_sentence(sp_sentence).encode('utf-8'))
<start> may i borrow this book ? <end> b'<start> \xc2\xbf puedo tomar prestado este libro ? <end>'
# 1. Remove the accents
# 2. Clean the sentences
# 3. Return word pairs in the format: [ENGLISH, SPANISH]
def create_dataset(path, num_examples):
lines = io.open(path, encoding='UTF-8').read().strip().split('\n')
word_pairs = [[preprocess_sentence(w) for w in l.split('\t')] for l in lines[:num_examples]]
return zip(*word_pairs)
en, sp = create_dataset(path_to_file, None)
print(en[-1])
print(sp[-1])
<start> if you want to sound like a native speaker , you must be willing to practice saying the same sentence over and over in the same way that banjo players practice the same phrase over and over until they can play it correctly and at the desired tempo . <end> <start> si quieres sonar como un hablante nativo , debes estar dispuesto a practicar diciendo la misma frase una y otra vez de la misma manera en que un musico de banjo practica el mismo fraseo una y otra vez hasta que lo puedan tocar correctamente y en el tiempo esperado . <end>
def tokenize(lang):
lang_tokenizer = tf.keras.preprocessing.text.Tokenizer(
filters='')
lang_tokenizer.fit_on_texts(lang)
tensor = lang_tokenizer.texts_to_sequences(lang)
tensor = tf.keras.preprocessing.sequence.pad_sequences(tensor,
padding='post')
return tensor, lang_tokenizer
def load_dataset(path, num_examples=None):
# creating cleaned input, output pairs
targ_lang, inp_lang = create_dataset(path, num_examples)
input_tensor, inp_lang_tokenizer = tokenize(inp_lang)
target_tensor, targ_lang_tokenizer = tokenize(targ_lang)
return input_tensor, target_tensor, inp_lang_tokenizer, targ_lang_tokenizer
Batasi ukuran dataset untuk bereksperimen lebih cepat (opsional)
Pelatihan tentang dataset lengkap> 100.000 kalimat akan memakan waktu lama. Untuk melatih lebih cepat, kami dapat membatasi ukuran dataset hingga 30.000 kalimat (tentu saja, kualitas terjemahan menurun dengan lebih sedikit data):
# Try experimenting with the size of that dataset
num_examples = 30000
input_tensor, target_tensor, inp_lang, targ_lang = load_dataset(path_to_file, num_examples)
# Calculate max_length of the target tensors
max_length_targ, max_length_inp = target_tensor.shape[1], input_tensor.shape[1]
# Creating training and validation sets using an 80-20 split
input_tensor_train, input_tensor_val, target_tensor_train, target_tensor_val = train_test_split(input_tensor, target_tensor, test_size=0.2)
# Show length
print(len(input_tensor_train), len(target_tensor_train), len(input_tensor_val), len(target_tensor_val))
24000 24000 6000 6000
def convert(lang, tensor):
for t in tensor:
if t!=0:
print ("%d ----> %s" % (t, lang.index_word[t]))
print ("Input Language; index to word mapping")
convert(inp_lang, input_tensor_train[0])
print ()
print ("Target Language; index to word mapping")
convert(targ_lang, target_tensor_train[0])
Input Language; index to word mapping 1 ----> <start> 4 ----> tom 42 ----> tiene 2344 ----> tos 3 ----> . 2 ----> <end> Target Language; index to word mapping 1 ----> <start> 5 ----> tom 51 ----> has 9 ----> a 1554 ----> cough 3 ----> . 2 ----> <end>
Buat dataset tf.data
BUFFER_SIZE = len(input_tensor_train)
BATCH_SIZE = 64
steps_per_epoch = len(input_tensor_train)//BATCH_SIZE
embedding_dim = 256
units = 1024
vocab_inp_size = len(inp_lang.word_index)+1
vocab_tar_size = len(targ_lang.word_index)+1
dataset = tf.data.Dataset.from_tensor_slices((input_tensor_train, target_tensor_train)).shuffle(BUFFER_SIZE)
dataset = dataset.batch(BATCH_SIZE, drop_remainder=True)
example_input_batch, example_target_batch = next(iter(dataset))
example_input_batch.shape, example_target_batch.shape
(TensorShape([64, 16]), TensorShape([64, 11]))
Tulis model encoder dan decoder
Menerapkan model encoder-decoder dengan perhatian yang dapat Anda baca di tutorial Terjemahan Mesin TensorFlow Neural (seq2seq) . Contoh ini menggunakan set API yang lebih baru. Notebook ini mengimplementasikan persamaan perhatian dari tutorial seq2seq. Diagram berikut menunjukkan bahwa setiap kata input diberi bobot oleh mekanisme perhatian yang kemudian digunakan oleh decoder untuk memprediksi kata berikutnya dalam kalimat. Gambar dan formula di bawah ini adalah contoh mekanisme perhatian dari kertas Luong .
Input dimasukkan melalui model encoder yang memberi kami output encoder bentuk (batch_size, max_length, hidden_size) dan bentuk bentuk tersembunyi encoder (batch_size, hidden_size) .
Berikut adalah persamaan yang diterapkan:
Tutorial ini menggunakan perhatian Bahdanau untuk pembuat enkode. Mari kita putuskan notasi sebelum menulis formulir yang disederhanakan:
- FC = Lapisan sepenuhnya terhubung (padat)
- EO = Output encoder
- H = keadaan tersembunyi
- X = input ke decoder
Dan pseudo-code:
-
score = FC(tanh(FC(EO) + FC(H))) -
attention weights = softmax(score, axis = 1). Softmax secara default diterapkan pada sumbu terakhir tetapi di sini kami ingin menerapkannya pada sumbu 1 , karena bentuk skornya adalah (batch_size, max_length, hidden_size) .Max_lengthadalah panjang input kami. Karena kami mencoba menetapkan bobot untuk setiap input, softmax harus diterapkan pada sumbu itu. -
context vector = sum(attention weights * EO, axis = 1). Alasan yang sama seperti di atas untuk memilih sumbu sebagai 1. -
embedding output= Input ke decoder X dilewatkan melalui lapisan embedding. -
merged vector = concat(embedding output, context vector) - Vektor gabungan ini kemudian diberikan kepada GRU
Bentuk semua vektor pada setiap langkah telah ditentukan dalam komentar dalam kode:
class Encoder(tf.keras.Model):
def __init__(self, vocab_size, embedding_dim, enc_units, batch_sz):
super(Encoder, self).__init__()
self.batch_sz = batch_sz
self.enc_units = enc_units
self.embedding = tf.keras.layers.Embedding(vocab_size, embedding_dim)
self.gru = tf.keras.layers.GRU(self.enc_units,
return_sequences=True,
return_state=True,
recurrent_initializer='glorot_uniform')
def call(self, x, hidden):
x = self.embedding(x)
output, state = self.gru(x, initial_state = hidden)
return output, state
def initialize_hidden_state(self):
return tf.zeros((self.batch_sz, self.enc_units))
encoder = Encoder(vocab_inp_size, embedding_dim, units, BATCH_SIZE)
# sample input
sample_hidden = encoder.initialize_hidden_state()
sample_output, sample_hidden = encoder(example_input_batch, sample_hidden)
print ('Encoder output shape: (batch size, sequence length, units) {}'.format(sample_output.shape))
print ('Encoder Hidden state shape: (batch size, units) {}'.format(sample_hidden.shape))
Encoder output shape: (batch size, sequence length, units) (64, 16, 1024) Encoder Hidden state shape: (batch size, units) (64, 1024)
class BahdanauAttention(tf.keras.layers.Layer):
def __init__(self, units):
super(BahdanauAttention, self).__init__()
self.W1 = tf.keras.layers.Dense(units)
self.W2 = tf.keras.layers.Dense(units)
self.V = tf.keras.layers.Dense(1)
def call(self, query, values):
# query hidden state shape == (batch_size, hidden size)
# query_with_time_axis shape == (batch_size, 1, hidden size)
# values shape == (batch_size, max_len, hidden size)
# we are doing this to broadcast addition along the time axis to calculate the score
query_with_time_axis = tf.expand_dims(query, 1)
# score shape == (batch_size, max_length, 1)
# we get 1 at the last axis because we are applying score to self.V
# the shape of the tensor before applying self.V is (batch_size, max_length, units)
score = self.V(tf.nn.tanh(
self.W1(query_with_time_axis) + self.W2(values)))
# attention_weights shape == (batch_size, max_length, 1)
attention_weights = tf.nn.softmax(score, axis=1)
# context_vector shape after sum == (batch_size, hidden_size)
context_vector = attention_weights * values
context_vector = tf.reduce_sum(context_vector, axis=1)
return context_vector, attention_weights
attention_layer = BahdanauAttention(10)
attention_result, attention_weights = attention_layer(sample_hidden, sample_output)
print("Attention result shape: (batch size, units) {}".format(attention_result.shape))
print("Attention weights shape: (batch_size, sequence_length, 1) {}".format(attention_weights.shape))
Attention result shape: (batch size, units) (64, 1024) Attention weights shape: (batch_size, sequence_length, 1) (64, 16, 1)
class Decoder(tf.keras.Model):
def __init__(self, vocab_size, embedding_dim, dec_units, batch_sz):
super(Decoder, self).__init__()
self.batch_sz = batch_sz
self.dec_units = dec_units
self.embedding = tf.keras.layers.Embedding(vocab_size, embedding_dim)
self.gru = tf.keras.layers.GRU(self.dec_units,
return_sequences=True,
return_state=True,
recurrent_initializer='glorot_uniform')
self.fc = tf.keras.layers.Dense(vocab_size)
# used for attention
self.attention = BahdanauAttention(self.dec_units)
def call(self, x, hidden, enc_output):
# enc_output shape == (batch_size, max_length, hidden_size)
context_vector, attention_weights = self.attention(hidden, enc_output)
# x shape after passing through embedding == (batch_size, 1, embedding_dim)
x = self.embedding(x)
# x shape after concatenation == (batch_size, 1, embedding_dim + hidden_size)
x = tf.concat([tf.expand_dims(context_vector, 1), x], axis=-1)
# passing the concatenated vector to the GRU
output, state = self.gru(x)
# output shape == (batch_size * 1, hidden_size)
output = tf.reshape(output, (-1, output.shape[2]))
# output shape == (batch_size, vocab)
x = self.fc(output)
return x, state, attention_weights
decoder = Decoder(vocab_tar_size, embedding_dim, units, BATCH_SIZE)
sample_decoder_output, _, _ = decoder(tf.random.uniform((BATCH_SIZE, 1)),
sample_hidden, sample_output)
print ('Decoder output shape: (batch_size, vocab size) {}'.format(sample_decoder_output.shape))
Decoder output shape: (batch_size, vocab size) (64, 4935)
Tentukan pengoptimal dan fungsi kerugian
optimizer = tf.keras.optimizers.Adam()
loss_object = tf.keras.losses.SparseCategoricalCrossentropy(
from_logits=True, reduction='none')
def loss_function(real, pred):
mask = tf.math.logical_not(tf.math.equal(real, 0))
loss_ = loss_object(real, pred)
mask = tf.cast(mask, dtype=loss_.dtype)
loss_ *= mask
return tf.reduce_mean(loss_)
Pos pemeriksaan (Penghematan berbasis objek)
checkpoint_dir = './training_checkpoints'
checkpoint_prefix = os.path.join(checkpoint_dir, "ckpt")
checkpoint = tf.train.Checkpoint(optimizer=optimizer,
encoder=encoder,
decoder=decoder)
Latihan
- Lewati input melalui encoder yang mengembalikan output encoder dan status tersembunyi encoder .
- Output encoder, keadaan tersembunyi encoder dan input decoder (yang merupakan token awal ) diteruskan ke decoder.
- Decoder mengembalikan prediksi dan status tersembunyi decoder .
- Status tersembunyi decoder kemudian dilewatkan kembali ke dalam model dan prediksi digunakan untuk menghitung kerugian.
- Gunakan memaksa guru untuk memutuskan input selanjutnya ke decoder.
- Guru memaksa adalah teknik di mana kata target dilewatkan sebagai input selanjutnya ke decoder.
- Langkah terakhir adalah menghitung gradien dan menerapkannya ke optimizer dan backpropagate.
@tf.function
def train_step(inp, targ, enc_hidden):
loss = 0
with tf.GradientTape() as tape:
enc_output, enc_hidden = encoder(inp, enc_hidden)
dec_hidden = enc_hidden
dec_input = tf.expand_dims([targ_lang.word_index['<start>']] * BATCH_SIZE, 1)
# Teacher forcing - feeding the target as the next input
for t in range(1, targ.shape[1]):
# passing enc_output to the decoder
predictions, dec_hidden, _ = decoder(dec_input, dec_hidden, enc_output)
loss += loss_function(targ[:, t], predictions)
# using teacher forcing
dec_input = tf.expand_dims(targ[:, t], 1)
batch_loss = (loss / int(targ.shape[1]))
variables = encoder.trainable_variables + decoder.trainable_variables
gradients = tape.gradient(loss, variables)
optimizer.apply_gradients(zip(gradients, variables))
return batch_loss
EPOCHS = 10
for epoch in range(EPOCHS):
start = time.time()
enc_hidden = encoder.initialize_hidden_state()
total_loss = 0
for (batch, (inp, targ)) in enumerate(dataset.take(steps_per_epoch)):
batch_loss = train_step(inp, targ, enc_hidden)
total_loss += batch_loss
if batch % 100 == 0:
print('Epoch {} Batch {} Loss {:.4f}'.format(epoch + 1,
batch,
batch_loss.numpy()))
# saving (checkpoint) the model every 2 epochs
if (epoch + 1) % 2 == 0:
checkpoint.save(file_prefix = checkpoint_prefix)
print('Epoch {} Loss {:.4f}'.format(epoch + 1,
total_loss / steps_per_epoch))
print('Time taken for 1 epoch {} sec\n'.format(time.time() - start))
Epoch 1 Batch 0 Loss 4.4937 Epoch 1 Batch 100 Loss 2.3472 Epoch 1 Batch 200 Loss 1.9153 Epoch 1 Batch 300 Loss 1.8042 Epoch 1 Loss 2.0265 Time taken for 1 epoch 27.345187664031982 sec Epoch 2 Batch 0 Loss 1.5260 Epoch 2 Batch 100 Loss 1.5228 Epoch 2 Batch 200 Loss 1.3840 Epoch 2 Batch 300 Loss 1.3131 Epoch 2 Loss 1.3900 Time taken for 1 epoch 15.777411222457886 sec Epoch 3 Batch 0 Loss 1.0458 Epoch 3 Batch 100 Loss 0.9216 Epoch 3 Batch 200 Loss 0.9254 Epoch 3 Batch 300 Loss 0.9041 Epoch 3 Loss 0.9699 Time taken for 1 epoch 15.391497373580933 sec Epoch 4 Batch 0 Loss 0.7582 Epoch 4 Batch 100 Loss 0.7201 Epoch 4 Batch 200 Loss 0.6765 Epoch 4 Batch 300 Loss 0.6696 Epoch 4 Loss 0.6555 Time taken for 1 epoch 15.782341480255127 sec Epoch 5 Batch 0 Loss 0.3534 Epoch 5 Batch 100 Loss 0.4191 Epoch 5 Batch 200 Loss 0.5322 Epoch 5 Batch 300 Loss 0.4767 Epoch 5 Loss 0.4494 Time taken for 1 epoch 15.508086204528809 sec Epoch 6 Batch 0 Loss 0.2508 Epoch 6 Batch 100 Loss 0.3366 Epoch 6 Batch 200 Loss 0.2935 Epoch 6 Batch 300 Loss 0.3432 Epoch 6 Loss 0.3137 Time taken for 1 epoch 15.811218738555908 sec Epoch 7 Batch 0 Loss 0.1759 Epoch 7 Batch 100 Loss 0.1997 Epoch 7 Batch 200 Loss 0.2879 Epoch 7 Batch 300 Loss 0.2643 Epoch 7 Loss 0.2257 Time taken for 1 epoch 15.454826831817627 sec Epoch 8 Batch 0 Loss 0.1318 Epoch 8 Batch 100 Loss 0.1151 Epoch 8 Batch 200 Loss 0.2130 Epoch 8 Batch 300 Loss 0.1852 Epoch 8 Loss 0.1712 Time taken for 1 epoch 15.786991596221924 sec Epoch 9 Batch 0 Loss 0.0876 Epoch 9 Batch 100 Loss 0.1227 Epoch 9 Batch 200 Loss 0.1361 Epoch 9 Batch 300 Loss 0.1682 Epoch 9 Loss 0.1328 Time taken for 1 epoch 15.443743467330933 sec Epoch 10 Batch 0 Loss 0.1048 Epoch 10 Batch 100 Loss 0.0736 Epoch 10 Batch 200 Loss 0.1056 Epoch 10 Batch 300 Loss 0.1204 Epoch 10 Loss 0.1074 Time taken for 1 epoch 15.615742683410645 sec
Menterjemahkan
- Fungsi evaluasi mirip dengan loop pelatihan, kecuali kami tidak menggunakan memaksa guru di sini. Input ke decoder pada setiap langkah waktu adalah prediksi sebelumnya bersama dengan status tersembunyi dan output encoder.
- Berhenti memprediksi ketika model memprediksi token akhir .
- Dan simpan bobot perhatian untuk setiap langkah waktu .
def evaluate(sentence):
attention_plot = np.zeros((max_length_targ, max_length_inp))
sentence = preprocess_sentence(sentence)
inputs = [inp_lang.word_index[i] for i in sentence.split(' ')]
inputs = tf.keras.preprocessing.sequence.pad_sequences([inputs],
maxlen=max_length_inp,
padding='post')
inputs = tf.convert_to_tensor(inputs)
result = ''
hidden = [tf.zeros((1, units))]
enc_out, enc_hidden = encoder(inputs, hidden)
dec_hidden = enc_hidden
dec_input = tf.expand_dims([targ_lang.word_index['<start>']], 0)
for t in range(max_length_targ):
predictions, dec_hidden, attention_weights = decoder(dec_input,
dec_hidden,
enc_out)
# storing the attention weights to plot later on
attention_weights = tf.reshape(attention_weights, (-1, ))
attention_plot[t] = attention_weights.numpy()
predicted_id = tf.argmax(predictions[0]).numpy()
result += targ_lang.index_word[predicted_id] + ' '
if targ_lang.index_word[predicted_id] == '<end>':
return result, sentence, attention_plot
# the predicted ID is fed back into the model
dec_input = tf.expand_dims([predicted_id], 0)
return result, sentence, attention_plot
# function for plotting the attention weights
def plot_attention(attention, sentence, predicted_sentence):
fig = plt.figure(figsize=(10,10))
ax = fig.add_subplot(1, 1, 1)
ax.matshow(attention, cmap='viridis')
fontdict = {'fontsize': 14}
ax.set_xticklabels([''] + sentence, fontdict=fontdict, rotation=90)
ax.set_yticklabels([''] + predicted_sentence, fontdict=fontdict)
ax.xaxis.set_major_locator(ticker.MultipleLocator(1))
ax.yaxis.set_major_locator(ticker.MultipleLocator(1))
plt.show()
def translate(sentence):
result, sentence, attention_plot = evaluate(sentence)
print('Input: %s' % (sentence))
print('Predicted translation: {}'.format(result))
attention_plot = attention_plot[:len(result.split(' ')), :len(sentence.split(' '))]
plot_attention(attention_plot, sentence.split(' '), result.split(' '))
Kembalikan pos pemeriksaan dan tes terbaru
# restoring the latest checkpoint in checkpoint_dir
checkpoint.restore(tf.train.latest_checkpoint(checkpoint_dir))
<tensorflow.python.training.tracking.util.CheckpointLoadStatus at 0x7ff366a4bcc0>
translate(u'hace mucho frio aqui.')
Input: <start> hace mucho frio aqui . <end> Predicted translation: it s very cold here . <end>
translate(u'esta es mi vida.')
Input: <start> esta es mi vida . <end> Predicted translation: this is my life . <end>
translate(u'¿todavia estan en casa?')
Input: <start> ¿ todavia estan en casa ? <end> Predicted translation: are you still at home ? <end>
# wrong translation
translate(u'trata de averiguarlo.')
Input: <start> trata de averiguarlo . <end> Predicted translation: try to figure it out . <end>
Langkah selanjutnya
- Unduh dataset berbeda untuk bereksperimen dengan terjemahan, misalnya, Bahasa Inggris ke Bahasa Jerman, atau Bahasa Inggris ke Bahasa Prancis.
- Eksperimen dengan pelatihan pada dataset yang lebih besar, atau menggunakan lebih banyak zaman