RNN によるテキスト生成

コレクションでコンテンツを整理 必要に応じて、コンテンツの保存と分類を行います。

QUEENE:
I had thought thou hadst a Roman; for the oracle,
Thus by All bids the man against the word,
Which are so weak of care, by old care done;
Your children were in your holy love,
And the precipitation through the bleeding throne.

BISHOP OF ELY:
Marry, and will, my lord, to weep in such a one were prettiest;
Yet now I was adopted heir
Of the world's lamentable day,
To watch the next way with his father with his face?

ESCALUS:
The cause why then we are all resolved more sons.

VOLUMNIA:
O, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, it is no sin it should be dead,
And love and pale as any will to that word.

QUEEN ELIZABETH:
But how long have I heard the soul for this world,
And show his hands of life be proved to stand.

PETRUCHIO:
I say he look'd on, if I must be content
To stay him from the fatal of our country's bliss.
His lordship pluck'd from this sentence then for prey,
And then let us twain, being the moon,
were she such a case as fills m

import tensorflow as tf

import numpy as np
import os
import time

2022-12-14 23:58:15.841952: W tensorflow/compiler/xla/stream_executor/platform/default/dso_loader.cc:64] Could not load dynamic library 'libnvinfer.so.7'; dlerror: libnvinfer.so.7: cannot open shared object file: No such file or directory
2022-12-14 23:58:15.842046: W tensorflow/compiler/xla/stream_executor/platform/default/dso_loader.cc:64] Could not load dynamic library 'libnvinfer_plugin.so.7'; dlerror: libnvinfer_plugin.so.7: cannot open shared object file: No such file or directory
2022-12-14 23:58:15.842056: W tensorflow/compiler/tf2tensorrt/utils/py_utils.cc:38] TF-TRT Warning: Cannot dlopen some TensorRT libraries. If you would like to use Nvidia GPU with TensorRT, please make sure the missing libraries mentioned above are installed properly.

path_to_file = tf.keras.utils.get_file('shakespeare.txt', 'https://storage.googleapis.com/download.tensorflow.org/data/shakespeare.txt')

# 読み込んだのち、Python 2 との互換性のためにデコード
text = open(path_to_file, 'rb').read().decode(encoding='utf-8')
# テキストの長さは含まれる文字数
print ('Length of text: {} characters'.format(len(text)))

Length of text: 1115394 characters

# テキストの最初の 250文字を参照
print(text[:250])

First Citizen:
Before we proceed any further, hear me speak.

All:
Speak, speak.

First Citizen:
You are all resolved rather to die than to famish?

All:
Resolved. resolved.

First Citizen:
First, you know Caius Marcius is chief enemy to the people.

# ファイル中のユニークな文字の数
vocab = sorted(set(text))
print ('{} unique characters'.format(len(vocab)))

65 unique characters

# それぞれの文字からインデックスへの対応表を作成
char2idx = {u:i for i, u in enumerate(vocab)}
idx2char = np.array(vocab)

text_as_int = np.array([char2idx[c] for c in text])

print('{')
for char,_ in zip(char2idx, range(20)):
    print('  {:4s}: {:3d},'.format(repr(char), char2idx[char]))
print('  ...\n}')

{
  '\n':   0,
  ' ' :   1,
  '!' :   2,
  '$' :   3,
  '&' :   4,
  "'" :   5,
  ',' :   6,
  '-' :   7,
  '.' :   8,
  '3' :   9,
  ':' :  10,
  ';' :  11,
  '?' :  12,
  'A' :  13,
  'B' :  14,
  'C' :  15,
  'D' :  16,
  'E' :  17,
  'F' :  18,
  'G' :  19,
  ...
}

# テキストの最初の 13 文字がどのように整数に変換されるかを見てみる
print ('{} ---- characters mapped to int ---- > {}'.format(repr(text[:13]), text_as_int[:13]))

'First Citizen' ---- characters mapped to int ---- > [18 47 56 57 58  1 15 47 58 47 64 43 52]

# ひとつの入力としたいシーケンスの文字数としての最大の長さ
seq_length = 100
examples_per_epoch = len(text)//(seq_length+1)

# 訓練用サンプルとターゲットを作る
char_dataset = tf.data.Dataset.from_tensor_slices(text_as_int)

for i in char_dataset.take(5):
  print(idx2char[i.numpy()])

F
i
r
s
t

sequences = char_dataset.batch(seq_length+1, drop_remainder=True)

for item in sequences.take(5):
  print(repr(''.join(idx2char[item.numpy()])))

'First Citizen:\nBefore we proceed any further, hear me speak.\n\nAll:\nSpeak, speak.\n\nFirst Citizen:\nYou '
'are all resolved rather to die than to famish?\n\nAll:\nResolved. resolved.\n\nFirst Citizen:\nFirst, you k'
"now Caius Marcius is chief enemy to the people.\n\nAll:\nWe know't, we know't.\n\nFirst Citizen:\nLet us ki"
"ll him, and we'll have corn at our own price.\nIs't a verdict?\n\nAll:\nNo more talking on't; let it be d"
'one: away, away!\n\nSecond Citizen:\nOne word, good citizens.\n\nFirst Citizen:\nWe are accounted poor citi'

def split_input_target(chunk):
    input_text = chunk[:-1]
    target_text = chunk[1:]
    return input_text, target_text

dataset = sequences.map(split_input_target)

for input_example, target_example in  dataset.take(1):
  print ('Input data: ', repr(''.join(idx2char[input_example.numpy()])))
  print ('Target data:', repr(''.join(idx2char[target_example.numpy()])))

Input data:  'First Citizen:\nBefore we proceed any further, hear me speak.\n\nAll:\nSpeak, speak.\n\nFirst Citizen:\nYou'
Target data: 'irst Citizen:\nBefore we proceed any further, hear me speak.\n\nAll:\nSpeak, speak.\n\nFirst Citizen:\nYou '

for i, (input_idx, target_idx) in enumerate(zip(input_example[:5], target_example[:5])):
    print("Step {:4d}".format(i))
    print("  input: {} ({:s})".format(input_idx, repr(idx2char[input_idx])))
    print("  expected output: {} ({:s})".format(target_idx, repr(idx2char[target_idx])))

Step    0
  input: 18 ('F')
  expected output: 47 ('i')
Step    1
  input: 47 ('i')
  expected output: 56 ('r')
Step    2
  input: 56 ('r')
  expected output: 57 ('s')
Step    3
  input: 57 ('s')
  expected output: 58 ('t')
Step    4
  input: 58 ('t')
  expected output: 1 (' ')

# バッチサイズ
BATCH_SIZE = 64

# データセットをシャッフルするためのバッファサイズ
# （TF data は可能性として無限長のシーケンスでも使えるように設計されています。
# このため、シーケンス全体をメモリ内でシャッフルしようとはしません。
# その代わりに、要素をシャッフルするためのバッファを保持しています）
BUFFER_SIZE = 10000

dataset = dataset.shuffle(BUFFER_SIZE).batch(BATCH_SIZE, drop_remainder=True)

dataset

<BatchDataset element_spec=(TensorSpec(shape=(64, 100), dtype=tf.int64, name=None), TensorSpec(shape=(64, 100), dtype=tf.int64, name=None))>

# 文字数で表されるボキャブラリーの長さ
vocab_size = len(vocab)

# 埋め込みベクトルの次元
embedding_dim = 256

# RNN ユニットの数
rnn_units = 1024

def build_model(vocab_size, embedding_dim, rnn_units, batch_size):
  model = tf.keras.Sequential([
    tf.keras.layers.Embedding(vocab_size, embedding_dim,
                              batch_input_shape=[batch_size, None]),
    tf.keras.layers.GRU(rnn_units,
                        return_sequences=True,
                        stateful=True,
                        recurrent_initializer='glorot_uniform'),
    tf.keras.layers.Dense(vocab_size)
  ])
  return model

model = build_model(
  vocab_size = len(vocab),
  embedding_dim=embedding_dim,
  rnn_units=rnn_units,
  batch_size=BATCH_SIZE)

for input_example_batch, target_example_batch in dataset.take(1):
  example_batch_predictions = model(input_example_batch)
  print(example_batch_predictions.shape, "# (batch_size, sequence_length, vocab_size)")

(64, 100, 65) # (batch_size, sequence_length, vocab_size)

model.summary()

Model: "sequential"
_________________________________________________________________
 Layer (type)                Output Shape              Param #   
=================================================================
 embedding (Embedding)       (64, None, 256)           16640     
                                                                 
 gru (GRU)                   (64, None, 1024)          3938304   
                                                                 
 dense (Dense)               (64, None, 65)            66625     
                                                                 
=================================================================
Total params: 4,021,569
Trainable params: 4,021,569
Non-trainable params: 0
_________________________________________________________________

sampled_indices = tf.random.categorical(example_batch_predictions[0], num_samples=1)
sampled_indices = tf.squeeze(sampled_indices,axis=-1).numpy()

sampled_indices

array([29,  4, 59, 43,  5, 52, 39,  4, 29,  6, 21, 50, 32, 36, 16,  5, 60,
        0, 43, 55, 25,  3, 14, 15, 39, 35,  6,  5, 22,  0, 44, 33,  1, 29,
       43, 17,  5, 54, 38,  6, 48, 40, 16, 49, 53, 15, 29, 51,  9,  8, 26,
       51, 60, 12,  6, 35, 12, 18, 52, 52, 45, 54, 13, 42, 12, 15, 19,  6,
       45,  7, 63,  7, 44, 23,  3, 26, 10, 54, 62, 36, 54, 22, 28,  5,  5,
       16, 32, 27, 57, 48,  4, 58,  0, 15, 34, 52, 35, 20, 50, 14])

print("Input: \n", repr("".join(idx2char[input_example_batch[0]])))
print()
print("Next Char Predictions: \n", repr("".join(idx2char[sampled_indices ])))

Input: 
 'uy and sell so, so give alms,\nPray so; and, for the ordering your affairs,\nTo sing them too: when yo'

Next Char Predictions: 
 "Q&ue'na&Q,IlTXD'v\neqM\\(BCaW,'J\nfU QeE'pZ,jbDkoCQm3.Nmv?,W?FnngpAd?CG,g-y-fK\\)N:pxXpJP''DTOsj&t\nCVnWHlB"

def loss(labels, logits):
  return tf.keras.losses.sparse_categorical_crossentropy(labels, logits, from_logits=True)

example_batch_loss  = loss(target_example_batch, example_batch_predictions)
print("Prediction shape: ", example_batch_predictions.shape, " # (batch_size, sequence_length, vocab_size)")
print("scalar_loss:      ", example_batch_loss.numpy().mean())

Prediction shape:  (64, 100, 65)  # (batch_size, sequence_length, vocab_size)
scalar_loss:       4.173783

model.compile(optimizer='adam', loss=loss)

# チェックポイントが保存されるディレクトリ
checkpoint_dir = './training_checkpoints'
# チェックポイントファイルの名称
checkpoint_prefix = os.path.join(checkpoint_dir, "ckpt_{epoch}")

checkpoint_callback=tf.keras.callbacks.ModelCheckpoint(
    filepath=checkpoint_prefix,
    save_weights_only=True)

EPOCHS=10

history = model.fit(dataset, epochs=EPOCHS, callbacks=[checkpoint_callback])

Epoch 1/10
172/172 [==============================] - 10s 42ms/step - loss: 2.6739
Epoch 2/10
172/172 [==============================] - 8s 38ms/step - loss: 1.9617
Epoch 3/10
172/172 [==============================] - 8s 38ms/step - loss: 1.6931
Epoch 4/10
172/172 [==============================] - 8s 38ms/step - loss: 1.5456
Epoch 5/10
172/172 [==============================] - 8s 38ms/step - loss: 1.4564
Epoch 6/10
172/172 [==============================] - 8s 38ms/step - loss: 1.3969
Epoch 7/10
172/172 [==============================] - 8s 38ms/step - loss: 1.3516
Epoch 8/10
172/172 [==============================] - 8s 38ms/step - loss: 1.3123
Epoch 9/10
172/172 [==============================] - 8s 38ms/step - loss: 1.2789
Epoch 10/10
172/172 [==============================] - 8s 38ms/step - loss: 1.2475

tf.train.latest_checkpoint(checkpoint_dir)

'./training_checkpoints/ckpt_10'

model = build_model(vocab_size, embedding_dim, rnn_units, batch_size=1)

model.load_weights(tf.train.latest_checkpoint(checkpoint_dir))

model.build(tf.TensorShape([1, None]))

model.summary()

Model: "sequential_1"
_________________________________________________________________
 Layer (type)                Output Shape              Param #   
=================================================================
 embedding_1 (Embedding)     (1, None, 256)            16640     
                                                                 
 gru_1 (GRU)                 (1, None, 1024)           3938304   
                                                                 
 dense_1 (Dense)             (1, None, 65)             66625     
                                                                 
=================================================================
Total params: 4,021,569
Trainable params: 4,021,569
Non-trainable params: 0
_________________________________________________________________

def generate_text(model, start_string):
  # 評価ステップ（学習済みモデルを使ったテキスト生成）

  # 生成する文字数
  num_generate = 1000

  # 開始文字列を数値に変換（ベクトル化）
  input_eval = [char2idx[s] for s in start_string]
  input_eval = tf.expand_dims(input_eval, 0)

  # 結果を保存する空文字列
  text_generated = []

  # 低い temperature　は、より予測しやすいテキストをもたらし
  # 高い temperature は、より意外なテキストをもたらす
  # 実験により最適な設定を見つけること
  temperature = 1.0

  # ここではバッチサイズ　== 1
  model.reset_states()
  for i in range(num_generate):
      predictions = model(input_eval)
      # バッチの次元を削除
      predictions = tf.squeeze(predictions, 0)

      # カテゴリー分布をつかってモデルから返された文字を予測 
      predictions = predictions / temperature
      predicted_id = tf.random.categorical(predictions, num_samples=1)[-1,0].numpy()

      # 過去の隠れ状態とともに予測された文字をモデルへのつぎの入力として渡す
      input_eval = tf.expand_dims([predicted_id], 0)

      text_generated.append(idx2char[predicted_id])

  return (start_string + ''.join(text_generated))

print(generate_text(model, start_string=u"ROMEO: "))

ROMEO: I am a miss'd.
But sleep forth. Be,
To chose by gailting than thy daughter's mind
Good queen, my lord.

VINCENTIO:
No, I'll proceed.

SICINIUS:
Put out an egry cape
Would have the wars friar'dly,
Hath colouris contemplisans.

SEBASTIAN:
Ay, the time thou but well, not for't.
Both living fortune waste with it forswore,
Thanks, and all to thine eye,
We shall have, she was rund. Hil away?

AUTOLYCUS:
Here canst thou never being to use ie fest
beloved o' the court-world and weakness,--

FLORIZEL:
Scorn, fair wetch your good company: masters, call'd up,
As call you beheld, lord.

LARTIUS:
He should do giving the plishes; these fellows a man,
Nay, let me weep
Than come upon his majesty! By bright-advice they but a fool,
And fetch itsell landu-doom here alive the day Tyrrel,
God knoces him for a very more voices.
We say, a thousand-Face that ever I am about to.

Poold:
So lie, the lust, the pant murder? is therely earth.

CATESBY:
Is, palket chargely forget,
When I go; go, come, or a better d

model = build_model(
  vocab_size = len(vocab),
  embedding_dim=embedding_dim,
  rnn_units=rnn_units,
  batch_size=BATCH_SIZE)

WARNING:tensorflow:Detecting that an object or model or tf.train.Checkpoint is being deleted with unrestored values. See the following logs for the specific values in question. To silence these warnings, use `status.expect_partial()`. See https://www.tensorflow.org/api_docs/python/tf/train/Checkpoint#restorefor details about the status object returned by the restore function.
WARNING:tensorflow:Value in checkpoint could not be found in the restored object: (root).optimizer._iterations
WARNING:tensorflow:Value in checkpoint could not be found in the restored object: (root).optimizer._learning_rate
WARNING:tensorflow:Value in checkpoint could not be found in the restored object: (root).optimizer._variables.1
WARNING:tensorflow:Value in checkpoint could not be found in the restored object: (root).optimizer._variables.2
WARNING:tensorflow:Value in checkpoint could not be found in the restored object: (root).optimizer._variables.3
WARNING:tensorflow:Value in checkpoint could not be found in the restored object: (root).optimizer._variables.4
WARNING:tensorflow:Value in checkpoint could not be found in the restored object: (root).optimizer._variables.5
WARNING:tensorflow:Value in checkpoint could not be found in the restored object: (root).optimizer._variables.6
WARNING:tensorflow:Value in checkpoint could not be found in the restored object: (root).optimizer._variables.7
WARNING:tensorflow:Value in checkpoint could not be found in the restored object: (root).optimizer._variables.8
WARNING:tensorflow:Value in checkpoint could not be found in the restored object: (root).optimizer._variables.9
WARNING:tensorflow:Value in checkpoint could not be found in the restored object: (root).optimizer._variables.10
WARNING:tensorflow:Value in checkpoint could not be found in the restored object: (root).optimizer._variables.11
WARNING:tensorflow:Value in checkpoint could not be found in the restored object: (root).optimizer._variables.12

optimizer = tf.keras.optimizers.Adam()

@tf.function
def train_step(inp, target):
  with tf.GradientTape() as tape:
    predictions = model(inp)
    loss = tf.reduce_mean(
        tf.keras.losses.sparse_categorical_crossentropy(
            target, predictions, from_logits=True))
  grads = tape.gradient(loss, model.trainable_variables)
  optimizer.apply_gradients(zip(grads, model.trainable_variables))

  return loss

# 訓練ステップ
EPOCHS = 10

for epoch in range(EPOCHS):
  start = time.time()

  # 各エポックの最初に、隠れ状態を初期化する
  # 最初は隠れ状態は None
  hidden = model.reset_states()

  for (batch_n, (inp, target)) in enumerate(dataset):
    loss = train_step(inp, target)

    if batch_n % 100 == 0:
      template = 'Epoch {} Batch {} Loss {}'
      print(template.format(epoch+1, batch_n, loss))

  # 5エポックごとにモデル（のチェックポイント）を保存する
  if (epoch + 1) % 5 == 0:
    model.save_weights(checkpoint_prefix.format(epoch=epoch))

  print ('Epoch {} Loss {:.4f}'.format(epoch+1, loss))
  print ('Time taken for 1 epoch {} sec\n'.format(time.time() - start))

model.save_weights(checkpoint_prefix.format(epoch=epoch))

Epoch 1 Batch 0 Loss 4.174540996551514
Epoch 1 Batch 100 Loss 2.370338201522827
Epoch 1 Loss 2.1595
Time taken for 1 epoch 9.26038932800293 sec

Epoch 2 Batch 0 Loss 2.1852879524230957
Epoch 2 Batch 100 Loss 1.9203600883483887
Epoch 2 Loss 1.8065
Time taken for 1 epoch 7.190315246582031 sec

Epoch 3 Batch 0 Loss 1.8183932304382324
Epoch 3 Batch 100 Loss 1.669272780418396
Epoch 3 Loss 1.6125
Time taken for 1 epoch 7.355443239212036 sec

Epoch 4 Batch 0 Loss 1.5592581033706665
Epoch 4 Batch 100 Loss 1.5509053468704224
Epoch 4 Loss 1.5413
Time taken for 1 epoch 7.222464084625244 sec

Epoch 5 Batch 0 Loss 1.4591556787490845
Epoch 5 Batch 100 Loss 1.470933198928833
Epoch 5 Loss 1.4629
Time taken for 1 epoch 7.2957847118377686 sec

Epoch 6 Batch 0 Loss 1.4055699110031128
Epoch 6 Batch 100 Loss 1.3990877866744995
Epoch 6 Loss 1.3728
Time taken for 1 epoch 7.287255764007568 sec

Epoch 7 Batch 0 Loss 1.3341516256332397
Epoch 7 Batch 100 Loss 1.3645275831222534
Epoch 7 Loss 1.3426
Time taken for 1 epoch 7.1413867473602295 sec

Epoch 8 Batch 0 Loss 1.2572031021118164
Epoch 8 Batch 100 Loss 1.3154652118682861
Epoch 8 Loss 1.2930
Time taken for 1 epoch 7.195223808288574 sec

Epoch 9 Batch 0 Loss 1.2512826919555664
Epoch 9 Batch 100 Loss 1.3123900890350342
Epoch 9 Loss 1.3117
Time taken for 1 epoch 7.14519190788269 sec

Epoch 10 Batch 0 Loss 1.1898956298828125
Epoch 10 Batch 100 Loss 1.2471652030944824
Epoch 10 Loss 1.2674
Time taken for 1 epoch 7.228768825531006 sec