RNN语言翻译
02 Jun 2018 | deep-learning |语言翻译
这个项目是使用RNN实现英语到法语的翻译。
"""
DON'T MODIFY ANYTHING IN THIS CELL
"""
import helper
import problem_unittests as tests
source_path = 'data/small_vocab_en'
target_path = 'data/small_vocab_fr'
source_text = helper.load_data(source_path)
target_text = helper.load_data(target_path)
查看数据
view_sentence_range = (0, 10)
"""
DON'T MODIFY ANYTHING IN THIS CELL
"""
import numpy as np
print('Dataset Stats')
print('Roughly the number of unique words: {}'.format(len({word: None for word in source_text.split()})))
sentences = source_text.split('\n')
word_counts = [len(sentence.split()) for sentence in sentences]
print('Number of sentences: {}'.format(len(sentences)))
print('Average number of words in a sentence: {}'.format(np.average(word_counts)))
print()
print('English sentences {} to {}:'.format(*))
print('\n'.join(source_text.split('\n')[view_sentence_range[0]:view_sentence_range[1]]))
print()
print('French sentences {} to {}:'.format(*view_sentence_range))
print('\n'.join(target_text.split('\n')[view_sentence_range[0]:view_sentence_range[1]]))
File "<ipython-input-2-f34785c0978f>", line 17
print('English sentences {} to {}:'.format(*))
^
SyntaxError: invalid syntax
预处理函数
文本转ID
与其他RNN一样，我们首先需要把单词转化为计算机能够处理的数值。函数text_to_ids()可以将文本转化为整数id，注意需要在末尾加上<EOS>，让神经网络知道一个句子的结束。
def text_to_ids(source_text, target_text, source_vocab_to_int, target_vocab_to_int):
"""
Convert source and target text to proper word ids
:param source_text: String that contains all the source text.
:param target_text: String that contains all the target text.
:param source_vocab_to_int: Dictionary to go from the source words to an id
:param target_vocab_to_int: Dictionary to go from the target words to an id
:return: A tuple of lists (source_id_text, target_id_text)
"""
# TODO: Implement Function
source_id_text = []
for sentences in source_text.split('\n'):
sentence_id_text = []
for word in sentences.split():
idx = source_vocab_to_int[word]
sentence_id_text.append(idx)
source_id_text.append(sentence_id_text)
target_id_text = []
for sentences in target_text.split('\n'):
sentence_id_text = []
for word in sentences.split():
idx = target_vocab_to_int[word]
sentence_id_text.append(idx)
sentence_id_text.append(target_vocab_to_int['<EOS>'])
target_id_text.append(sentence_id_text)
return (source_id_text, target_id_text)
"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
tests.test_text_to_ids(text_to_ids)
保存处理后的数据
"""
DON'T MODIFY ANYTHING IN THIS CELL
"""
helper.preprocess_and_save_data(source_path, target_path, text_to_ids)
"""
DON'T MODIFY ANYTHING IN THIS CELL
"""
import numpy as np
import helper
import problem_unittests as tests
(source_int_text, target_int_text), (source_vocab_to_int, target_vocab_to_int), _ = helper.load_preprocess()
查看TensorFlow的版本
"""
DON'T MODIFY ANYTHING IN THIS CELL
"""
from distutils.version import LooseVersion
import warnings
import tensorflow as tf
from tensorflow.python.layers.core import Dense
# Check TensorFlow Version
assert LooseVersion(tf.__version__) >= LooseVersion('1.1'), 'Please use TensorFlow version 1.1 or newer'
print('TensorFlow Version: {}'.format(tf.__version__))
# Check for a GPU
if not tf.test.gpu_device_name():
warnings.warn('No GPU found. Please use a GPU to train your neural network.')
else:
print('Default GPU Device: {}'.format(tf.test.gpu_device_name()))
TensorFlow Version: 1.1.0
d:\Anaconda3\envs\tf1.1\lib\site-packages\ipykernel_launcher.py:15: UserWarning: No GPU found. Please use a GPU to train your neural network.
from ipykernel import kernelapp as app
构建神经网络
为了构建序列到序列的神经网络模型，我们需要实现下列模块：
model_inputsprocess_decoder_inputencoding_layerdecoding_layer_traindecoding_layer_inferdecoding_layerseq2seq_model
Input
函数model_inputs()用于创建输入参数。
def model_inputs():
"""
Create TF Placeholders for input, targets, learning rate, and lengths of source and target sequences.
:return: Tuple (input, targets, learning rate, keep probability, target sequence length,
max target sequence length, source sequence length)
"""
# TODO: Implement Function
input_ = tf.placeholder(tf.int32, [None, None], name='input')
targets = tf.placeholder(tf.int32, [None, None], name='targets')
lr = tf.placeholder(tf.float32, name='learning_rate')
keep_prob = tf.placeholder(tf.float32, name='keep_prob')
target_sequence_length = tf.placeholder(tf.int32, (None, ), name='target_sequence_length')
max_target_len = tf.reduce_max(target_sequence_length, name='max_target_len')
source_sequence_length = tf.placeholder(tf.int32, (None, ), name='source_sequence_length')
return input_, targets, lr, keep_prob, target_sequence_length, max_target_len, source_sequence_length
"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
tests.test_model_inputs(model_inputs)
Tests Passed
解码输入
def process_decoder_input(target_data, target_vocab_to_int, batch_size):
"""
Preprocess target data for encoding
:param target_data: Target Placehoder
:param target_vocab_to_int: Dictionary to go from the target words to an id
:param batch_size: Batch Size
:return: Preprocessed target data
"""
# TODO: Implement Function
ending = tf.strided_slice(target_data, [0, 0], [batch_size, -1], [1,1] ) # 移除最后一个id
dec_input = tf.concat([tf.fill([batch_size, 1], target_vocab_to_int['<GO>']), ending ], 1) #添加开始id
return dec_input
"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
tests.test_process_encoding_input(process_decoder_input)
Tests Passed
Encoding
函数encoding_layer() 用于创建RNN的编码层:
- 使用
tf.contrib.layers.embed_sequence实现嵌入的编码
from imp import reload
reload(tests)
def encoding_layer(rnn_inputs, rnn_size, num_layers, keep_prob,
source_sequence_length, source_vocab_size,
encoding_embedding_size):
"""
Create encoding layer
:param rnn_inputs: Inputs for the RNN
:param rnn_size: RNN Size
:param num_layers: Number of layers
:param keep_prob: Dropout keep probability
:param source_sequence_length: a list of the lengths of each sequence in the batch
:param source_vocab_size: vocabulary size of source data
:param encoding_embedding_size: embedding size of source data
:return: tuple (RNN output, RNN state)
"""
# TODO: Implement Function
enc_embed_input = tf.contrib.layers.embed_sequence(rnn_inputs, source_vocab_size, encoding_embedding_size)# id to vector
#RNN Cell
def make_cell(rnn_size):
enc_cell = tf.contrib.rnn.LSTMCell(rnn_size, initializer=tf.random_normal_initializer(-0.1, 0.1, seed=2))
enc_cell = tf.contrib.rnn.DropoutWrapper(enc_cell, keep_prob)
return enc_cell
enc_cell = tf.contrib.rnn.MultiRNNCell([make_cell(rnn_size) for _ in range(num_layers)] )
enc_output, enc_state = tf.nn.dynamic_rnn(enc_cell, enc_embed_input, sequence_length=source_sequence_length, dtype=tf.float32)
return enc_output, enc_state
"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
tests.test_encoding_layer(encoding_layer)
Tests Passed
解码-训练
创建训练解码层:
- 训练：
tf.contrib.seq2seq.TrainingHelper - 编码：
tf.contrib.seq2seq.BasicDecoder - 输出：
tf.contrib.seq2seq.dynamic_decode
def decoding_layer_train(encoder_state, dec_cell, dec_embed_input,
target_sequence_length, max_summary_length,
output_layer, keep_prob):
"""
Create a decoding layer for training
:param encoder_state: Encoder State
:param dec_cell: Decoder RNN Cell
:param dec_embed_input: Decoder embedded input
:param target_sequence_length: The lengths of each sequence in the target batch
:param max_summary_length: The length of the longest sequence in the batch
:param output_layer: Function to apply the output layer
:param keep_prob: Dropout keep probability
:return: BasicDecoderOutput containing training logits and sample_id
"""
# TODO: Implement Function
trainHelper = tf.contrib.seq2seq.TrainingHelper(inputs=dec_embed_input, sequence_length=target_sequence_length, time_major=False)
train_decoder = tf.contrib.seq2seq.BasicDecoder(dec_cell, trainHelper, encoder_state, output_layer)
train_decoder_out = tf.contrib.seq2seq.dynamic_decode(train_decoder, impute_finished=True, maximum_iterations=max_summary_length)[0]
return train_decoder_out
"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
tests.test_decoding_layer_train(decoding_layer_train)
Tests Passed
编码-推断
创建推断解码层:
- 输入：
tf.contrib.seq2seq.GreedyEmbeddingHelper - 编码：
tf.contrib.seq2seq.BasicDecoder - 输出：
tf.contrib.seq2seq.dynamic_decode
def decoding_layer_infer(encoder_state, dec_cell, dec_embeddings, start_of_sequence_id,
end_of_sequence_id, max_target_sequence_length,
vocab_size, output_layer, batch_size, keep_prob):
"""
Create a decoding layer for inference
:param encoder_state: Encoder state
:param dec_cell: Decoder RNN Cell
:param dec_embeddings: Decoder embeddings
:param start_of_sequence_id: GO ID
:param end_of_sequence_id: EOS Id
:param max_target_sequence_length: Maximum length of target sequences
:param vocab_size: Size of decoder/target vocabulary
:param decoding_scope: TenorFlow Variable Scope for decoding
:param output_layer: Function to apply the output layer
:param batch_size: Batch size
:param keep_prob: Dropout keep probability
:return: BasicDecoderOutput containing inference logits and sample_id
"""
# TODO: Implement Function
#start_tile = tf.tile(tf.constant(start_of_sequence_id,dtype=tf.int32), [batch_size])
#start_tokens = np.array([start_of_sequence_id] * batch_size)[None:]
start_tokens = tf.tile(tf.constant([start_of_sequence_id], dtype=tf.int32), [batch_size], name='start_tokens')
inference_helper = tf.contrib.seq2seq.GreedyEmbeddingHelper(dec_embeddings, start_tokens, end_of_sequence_id)
inference_decoder = tf.contrib.seq2seq.BasicDecoder(dec_cell, inference_helper, encoder_state, output_layer)
inference_decoder_output = tf.contrib.seq2seq.dynamic_decode(inference_decoder,
impute_finished=True,
maximum_iterations=max_target_sequence_length)[0]
return inference_decoder_output
"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
tests.test_decoding_layer_infer(decoding_layer_infer)
Tests Passed
解码层
函数decoding_layer()创建RNN解码层.
注意: 使用tf.variable_scope在训练与解码是共享变量。
def decoding_layer(dec_input, encoder_state,
target_sequence_length, max_target_sequence_length,
rnn_size,
num_layers, target_vocab_to_int, target_vocab_size,
batch_size, keep_prob, decoding_embedding_size):
"""
Create decoding layer
:param dec_input: Decoder input
:param encoder_state: Encoder state
:param target_sequence_length: The lengths of each sequence in the target batch
:param max_target_sequence_length: Maximum length of target sequences
:param rnn_size: RNN Size
:param num_layers: Number of layers
:param target_vocab_to_int: Dictionary to go from the target words to an id
:param target_vocab_size: Size of target vocabulary
:param batch_size: The size of the batch
:param keep_prob: Dropout keep probability
:param decoding_embedding_size: Decoding embedding size
:return: Tuple of (Training BasicDecoderOutput, Inference BasicDecoderOutput)
"""
# TODO: Implement Function
# 1. building decoder_embedding
dec_embeddings = tf.Variable(tf.random_uniform([target_vocab_size, decoding_embedding_size]))
dec_embed_input = tf.nn.embedding_lookup(dec_embeddings, dec_input)
# 2. construct decoder cell
def make_cell(rnn_size):
dec_cell = tf.contrib.rnn.LSTMCell(rnn_size, initializer=tf.random_uniform_initializer(-0.1, 0.1, seed=2))
dec_cell = tf.contrib.rnn.DropoutWrapper(dec_cell, keep_prob)
return dec_cell
dec_cell = tf.contrib.rnn.MultiRNNCell([make_cell(rnn_size) for _ in range(num_layers)])
# 3. construct output layer
output_layer = Dense(target_vocab_size, kernel_initializer=tf.random_normal_initializer(mean=0.0, stddev=0.1))
# 4 & 5 get (Training BasicDecoderOutput, Inference BasicDecoderOutput)
with tf.variable_scope('decode'):
training_decoder_output = decoding_layer_train(encoder_state, dec_cell, dec_embed_input,
target_sequence_length, max_target_sequence_length, output_layer, keep_prob)
with tf.variable_scope('decode', reuse=True):
inference_decoder_output = decoding_layer_infer(encoder_state, dec_cell, dec_embeddings, target_vocab_to_int['<GO>'],
target_vocab_to_int['<EOS>'], max_target_sequence_length, target_vocab_size, output_layer, batch_size, keep_prob)
return training_decoder_output, inference_decoder_output
"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
tests.test_decoding_layer(decoding_layer)
Tests Passed
构建神经网络
def seq2seq_model(input_data, target_data, keep_prob, batch_size,
source_sequence_length, target_sequence_length,
max_target_sentence_length,
source_vocab_size, target_vocab_size,
enc_embedding_size, dec_embedding_size,
rnn_size, num_layers, target_vocab_to_int):
"""
Build the Sequence-to-Sequence part of the neural network
:param input_data: Input placeholder
:param target_data: Target placeholder
:param keep_prob: Dropout keep probability placeholder
:param batch_size: Batch Size
:param source_sequence_length: Sequence Lengths of source sequences in the batch
:param target_sequence_length: Sequence Lengths of target sequences in the batch
:param source_vocab_size: Source vocabulary size
:param target_vocab_size: Target vocabulary size
:param enc_embedding_size: Encoder embedding size
:param dec_embedding_size: Decoder embedding size
:param rnn_size: RNN Size
:param num_layers: Number of layers
:param target_vocab_to_int: Dictionary to go from the target words to an id
:return: Tuple of (Training BasicDecoderOutput, Inference BasicDecoderOutput)
"""
# TODO: Implement Function
_, enc_state = encoding_layer(input_data, rnn_size, num_layers, keep_prob,
source_sequence_length, source_vocab_size,
enc_embedding_size)
dec_input = process_decoder_input(target_data, target_vocab_to_int, batch_size)
train_decoder_output, inference_decoder_output = decoding_layer(dec_input, enc_state, target_sequence_length,
max_target_sentence_length, rnn_size,num_layers,
target_vocab_to_int, target_vocab_size, batch_size,
keep_prob, dec_embedding_size)
return train_decoder_output, inference_decoder_output
"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
tests.test_seq2seq_model(seq2seq_model)
Tests Passed
训练神经网络
超参数
# Number of Epochs
epochs = 50
# Batch Size
batch_size = 128
# RNN Size
rnn_size = 128
# Number of Layers
num_layers = 2
# Embedding Size
encoding_embedding_size = 200
decoding_embedding_size = 200
# Learning Rate
learning_rate = 0.001
# Dropout Keep Probability
keep_probability = 0.8
display_step = 5
构建图
"""
DON'T MODIFY ANYTHING IN THIS CELL
"""
save_path = 'checkpoints/dev'
(source_int_text, target_int_text), (source_vocab_to_int, target_vocab_to_int), _ = helper.load_preprocess()
max_target_sentence_length = max([len(sentence) for sentence in source_int_text])
train_graph = tf.Graph()
with train_graph.as_default():
input_data, targets, lr, keep_prob, target_sequence_length, max_target_sequence_length, source_sequence_length = model_inputs()
#sequence_length = tf.placeholder_with_default(max_target_sentence_length, None, name='sequence_length')
input_shape = tf.shape(input_data)
train_logits, inference_logits = seq2seq_model(tf.reverse(input_data, [-1]),
targets,
keep_prob,
batch_size,
source_sequence_length,
target_sequence_length,
max_target_sequence_length,
len(source_vocab_to_int),
len(target_vocab_to_int),
encoding_embedding_size,
decoding_embedding_size,
rnn_size,
num_layers,
target_vocab_to_int)
training_logits = tf.identity(train_logits.rnn_output, name='logits')
inference_logits = tf.identity(inference_logits.sample_id, name='predictions')
masks = tf.sequence_mask(target_sequence_length, max_target_sequence_length, dtype=tf.float32, name='masks')
with tf.name_scope("optimization"):
# Loss function
cost = tf.contrib.seq2seq.sequence_loss(
training_logits,
targets,
masks)
# Optimizer
optimizer = tf.train.AdamOptimizer(lr)
# Gradient Clipping
gradients = optimizer.compute_gradients(cost)
capped_gradients = [(tf.clip_by_value(grad, -1., 1.), var) for grad, var in gradients if grad is not None]
train_op = optimizer.apply_gradients(capped_gradients)
"""
DON'T MODIFY ANYTHING IN THIS CELL
"""
def pad_sentence_batch(sentence_batch, pad_int):
"""Pad sentences with <PAD> so that each sentence of a batch has the same length"""
max_sentence = max([len(sentence) for sentence in sentence_batch])
return [sentence + [pad_int] * (max_sentence - len(sentence)) for sentence in sentence_batch]
def get_batches(sources, targets, batch_size, source_pad_int, target_pad_int):
"""Batch targets, sources, and the lengths of their sentences together"""
for batch_i in range(0, len(sources)//batch_size):
start_i = batch_i * batch_size
# Slice the right amount for the batch
sources_batch = sources[start_i:start_i + batch_size]
targets_batch = targets[start_i:start_i + batch_size]
# Pad
pad_sources_batch = np.array(pad_sentence_batch(sources_batch, source_pad_int))
pad_targets_batch = np.array(pad_sentence_batch(targets_batch, target_pad_int))
# Need the lengths for the _lengths parameters
pad_targets_lengths = []
for target in pad_targets_batch:
pad_targets_lengths.append(len(target))
pad_source_lengths = []
for source in pad_sources_batch:
pad_source_lengths.append(len(source))
yield pad_sources_batch, pad_targets_batch, pad_source_lengths, pad_targets_lengths
"""
DON'T MODIFY ANYTHING IN THIS CELL
"""
def get_accuracy(target, logits):
"""
Calculate accuracy
"""
max_seq = max(target.shape[1], logits.shape[1])
if max_seq - target.shape[1]:
target = np.pad(
target,
[(0,0),(0,max_seq - target.shape[1])],
'constant')
if max_seq - logits.shape[1]:
logits = np.pad(
logits,
[(0,0),(0,max_seq - logits.shape[1])],
'constant')
return np.mean(np.equal(target, logits))
# Split data to training and validation sets
train_source = source_int_text[batch_size:]
train_target = target_int_text[batch_size:]
valid_source = source_int_text[:batch_size]
valid_target = target_int_text[:batch_size]
(valid_sources_batch, valid_targets_batch, valid_sources_lengths, valid_targets_lengths ) = next(get_batches(valid_source,
valid_target,
batch_size,
source_vocab_to_int['<PAD>'],
target_vocab_to_int['<PAD>']))
with tf.Session(graph=train_graph) as sess:
sess.run(tf.global_variables_initializer())
for epoch_i in range(epochs):
for batch_i, (source_batch, target_batch, sources_lengths, targets_lengths) in enumerate(
get_batches(train_source, train_target, batch_size,
source_vocab_to_int['<PAD>'],
target_vocab_to_int['<PAD>'])):
_, loss = sess.run(
[train_op, cost],
{input_data: source_batch,
targets: target_batch,
lr: learning_rate,
target_sequence_length: targets_lengths,
source_sequence_length: sources_lengths,
keep_prob: keep_probability})
if batch_i % display_step == 0 and batch_i > 0:
batch_train_logits = sess.run(
inference_logits,
{input_data: source_batch,
source_sequence_length: sources_lengths,
target_sequence_length: targets_lengths,
keep_prob: 1.0})
batch_valid_logits = sess.run(
inference_logits,
{input_data: valid_sources_batch,
source_sequence_length: valid_sources_lengths,
target_sequence_length: valid_targets_lengths,
keep_prob: 1.0})
train_acc = get_accuracy(target_batch, batch_train_logits)
valid_acc = get_accuracy(valid_targets_batch, batch_valid_logits)
print('Epoch {:>3} Batch {:>4}/{} - Train Accuracy: {:>6.4f}, Validation Accuracy: {:>6.4f}, Loss: {:>6.4f}'
.format(epoch_i, batch_i, len(source_int_text) // batch_size, train_acc, valid_acc, loss))
# Save Model
saver = tf.train.Saver()
saver.save(sess, save_path)
print('Model Trained and Saved')
Epoch 0 Batch 5/1077 - Train Accuracy: 0.2605, Validation Accuracy: 0.3050, Loss: 4.8083
Epoch 0 Batch 10/1077 - Train Accuracy: 0.1982, Validation Accuracy: 0.3050, Loss: 4.5230
Epoch 0 Batch 15/1077 - Train Accuracy: 0.2355, Validation Accuracy: 0.3050, Loss: 4.0101
Epoch 0 Batch 20/1077 - Train Accuracy: 0.2676, Validation Accuracy: 0.3349, Loss: 3.7757
Epoch 0 Batch 25/1077 - Train Accuracy: 0.2738, Validation Accuracy: 0.3363, Loss: 3.6803
Epoch 0 Batch 30/1077 - Train Accuracy: 0.2668, Validation Accuracy: 0.3366, Loss: 3.5269
Epoch 0 Batch 35/1077 - Train Accuracy: 0.2840, Validation Accuracy: 0.3452, Loss: 3.4180
Epoch 0 Batch 40/1077 - Train Accuracy: 0.2859, Validation Accuracy: 0.3540, Loss: 3.3374
Epoch 0 Batch 45/1077 - Train Accuracy: 0.2906, Validation Accuracy: 0.3558, Loss: 3.2612
Epoch 0 Batch 50/1077 - Train Accuracy: 0.2734, Validation Accuracy: 0.3398, Loss: 3.2286
Epoch 0 Batch 55/1077 - Train Accuracy: 0.3180, Validation Accuracy: 0.3633, Loss: 3.0543
Epoch 0 Batch 60/1077 - Train Accuracy: 0.3129, Validation Accuracy: 0.3548, Loss: 2.9829
Epoch 0 Batch 65/1077 - Train Accuracy: 0.2467, Validation Accuracy: 0.3452, Loss: 3.2078
Epoch 0 Batch 70/1077 - Train Accuracy: 0.2385, Validation Accuracy: 0.3434, Loss: 3.1128
Epoch 0 Batch 75/1077 - Train Accuracy: 0.2992, Validation Accuracy: 0.3580, Loss: 2.8538
Epoch 0 Batch 80/1077 - Train Accuracy: 0.3152, Validation Accuracy: 0.3757, Loss: 2.8607
Epoch 0 Batch 85/1077 - Train Accuracy: 0.3309, Validation Accuracy: 0.3885, Loss: 2.7357
Epoch 0 Batch 90/1077 - Train Accuracy: 0.3281, Validation Accuracy: 0.4020, Loss: 2.8434
Epoch 0 Batch 95/1077 - Train Accuracy: 0.3575, Validation Accuracy: 0.3981, Loss: 2.6422
Epoch 0 Batch 100/1077 - Train Accuracy: 0.3445, Validation Accuracy: 0.4059, Loss: 2.7220
Epoch 0 Batch 105/1077 - Train Accuracy: 0.3582, Validation Accuracy: 0.4094, Loss: 2.6024
Epoch 0 Batch 110/1077 - Train Accuracy: 0.3855, Validation Accuracy: 0.4240, Loss: 2.5641
Epoch 0 Batch 115/1077 - Train Accuracy: 0.3605, Validation Accuracy: 0.4268, Loss: 2.6026
Epoch 0 Batch 120/1077 - Train Accuracy: 0.3785, Validation Accuracy: 0.4268, Loss: 2.5283
Epoch 0 Batch 125/1077 - Train Accuracy: 0.3865, Validation Accuracy: 0.4414, Loss: 2.4286
Epoch 0 Batch 130/1077 - Train Accuracy: 0.4062, Validation Accuracy: 0.4602, Loss: 2.3033
Epoch 0 Batch 135/1077 - Train Accuracy: 0.3664, Validation Accuracy: 0.4528, Loss: 2.4416
Epoch 0 Batch 140/1077 - Train Accuracy: 0.3363, Validation Accuracy: 0.4318, Loss: 2.4486
Epoch 0 Batch 145/1077 - Train Accuracy: 0.3922, Validation Accuracy: 0.4347, Loss: 2.2174
Epoch 0 Batch 150/1077 - Train Accuracy: 0.3984, Validation Accuracy: 0.4403, Loss: 2.1143
Epoch 0 Batch 155/1077 - Train Accuracy: 0.3762, Validation Accuracy: 0.4389, Loss: 2.1498
Epoch 0 Batch 160/1077 - Train Accuracy: 0.4012, Validation Accuracy: 0.4389, Loss: 2.0739
Epoch 0 Batch 165/1077 - Train Accuracy: 0.3719, Validation Accuracy: 0.4453, Loss: 2.0320
Epoch 0 Batch 170/1077 - Train Accuracy: 0.3695, Validation Accuracy: 0.4457, Loss: 2.0678
Epoch 0 Batch 175/1077 - Train Accuracy: 0.3949, Validation Accuracy: 0.4428, Loss: 1.9046
Epoch 0 Batch 180/1077 - Train Accuracy: 0.4031, Validation Accuracy: 0.4474, Loss: 1.9053
Epoch 0 Batch 185/1077 - Train Accuracy: 0.3648, Validation Accuracy: 0.4421, Loss: 1.8736
Epoch 0 Batch 190/1077 - Train Accuracy: 0.4102, Validation Accuracy: 0.4513, Loss: 1.8367
Epoch 0 Batch 195/1077 - Train Accuracy: 0.3648, Validation Accuracy: 0.4421, Loss: 1.7875
Epoch 0 Batch 200/1077 - Train Accuracy: 0.3961, Validation Accuracy: 0.4471, Loss: 1.7693
Epoch 0 Batch 205/1077 - Train Accuracy: 0.3793, Validation Accuracy: 0.4442, Loss: 1.7260
Epoch 0 Batch 210/1077 - Train Accuracy: 0.3862, Validation Accuracy: 0.4421, Loss: 1.6790
Epoch 0 Batch 215/1077 - Train Accuracy: 0.3828, Validation Accuracy: 0.4460, Loss: 1.6478
Epoch 0 Batch 220/1077 - Train Accuracy: 0.3631, Validation Accuracy: 0.4439, Loss: 1.6524
Epoch 0 Batch 225/1077 - Train Accuracy: 0.3785, Validation Accuracy: 0.4567, Loss: 1.6018
Epoch 0 Batch 230/1077 - Train Accuracy: 0.4271, Validation Accuracy: 0.4418, Loss: 1.4768
Epoch 0 Batch 235/1077 - Train Accuracy: 0.4464, Validation Accuracy: 0.4496, Loss: 1.4412
Epoch 0 Batch 240/1077 - Train Accuracy: 0.4027, Validation Accuracy: 0.4492, Loss: 1.4963
Epoch 0 Batch 245/1077 - Train Accuracy: 0.4208, Validation Accuracy: 0.4570, Loss: 1.4102
Epoch 0 Batch 250/1077 - Train Accuracy: 0.4293, Validation Accuracy: 0.4496, Loss: 1.3399
Epoch 0 Batch 255/1077 - Train Accuracy: 0.3781, Validation Accuracy: 0.4496, Loss: 1.4565
Epoch 0 Batch 260/1077 - Train Accuracy: 0.4252, Validation Accuracy: 0.4570, Loss: 1.3361
Epoch 0 Batch 265/1077 - Train Accuracy: 0.4125, Validation Accuracy: 0.4492, Loss: 1.3508
Epoch 0 Batch 270/1077 - Train Accuracy: 0.3621, Validation Accuracy: 0.4542, Loss: 1.4343
Epoch 0 Batch 275/1077 - Train Accuracy: 0.4483, Validation Accuracy: 0.4609, Loss: 1.2768
Epoch 0 Batch 280/1077 - Train Accuracy: 0.4137, Validation Accuracy: 0.4513, Loss: 1.3412
Epoch 0 Batch 285/1077 - Train Accuracy: 0.4554, Validation Accuracy: 0.4638, Loss: 1.2315
Epoch 0 Batch 290/1077 - Train Accuracy: 0.4156, Validation Accuracy: 0.4542, Loss: 1.3239
Epoch 0 Batch 295/1077 - Train Accuracy: 0.4198, Validation Accuracy: 0.4769, Loss: 1.3635
Epoch 0 Batch 300/1077 - Train Accuracy: 0.3910, Validation Accuracy: 0.4769, Loss: 1.3149
Epoch 0 Batch 305/1077 - Train Accuracy: 0.4355, Validation Accuracy: 0.4691, Loss: 1.2292
Epoch 0 Batch 310/1077 - Train Accuracy: 0.4016, Validation Accuracy: 0.4773, Loss: 1.2079
Epoch 0 Batch 315/1077 - Train Accuracy: 0.4416, Validation Accuracy: 0.4776, Loss: 1.1580
Epoch 0 Batch 320/1077 - Train Accuracy: 0.4359, Validation Accuracy: 0.4826, Loss: 1.2018
Epoch 0 Batch 325/1077 - Train Accuracy: 0.4725, Validation Accuracy: 0.4727, Loss: 1.1155
Epoch 0 Batch 330/1077 - Train Accuracy: 0.4535, Validation Accuracy: 0.4759, Loss: 1.1712
Epoch 0 Batch 335/1077 - Train Accuracy: 0.5104, Validation Accuracy: 0.5025, Loss: 1.0513
Epoch 0 Batch 340/1077 - Train Accuracy: 0.4194, Validation Accuracy: 0.4954, Loss: 1.1615
Epoch 0 Batch 345/1077 - Train Accuracy: 0.4408, Validation Accuracy: 0.4801, Loss: 1.0666
Epoch 0 Batch 350/1077 - Train Accuracy: 0.4066, Validation Accuracy: 0.4936, Loss: 1.1759
Epoch 0 Batch 355/1077 - Train Accuracy: 0.4263, Validation Accuracy: 0.4844, Loss: 1.0969
Epoch 0 Batch 360/1077 - Train Accuracy: 0.4637, Validation Accuracy: 0.5249, Loss: 1.0746
Epoch 0 Batch 365/1077 - Train Accuracy: 0.4512, Validation Accuracy: 0.5241, Loss: 1.0831
Epoch 0 Batch 370/1077 - Train Accuracy: 0.4870, Validation Accuracy: 0.5234, Loss: 1.0119
Epoch 0 Batch 375/1077 - Train Accuracy: 0.5323, Validation Accuracy: 0.5270, Loss: 0.9655
Epoch 0 Batch 380/1077 - Train Accuracy: 0.4953, Validation Accuracy: 0.5412, Loss: 1.0100
Epoch 0 Batch 385/1077 - Train Accuracy: 0.5059, Validation Accuracy: 0.5341, Loss: 1.0293
Epoch 0 Batch 390/1077 - Train Accuracy: 0.4551, Validation Accuracy: 0.5295, Loss: 1.0598
Epoch 0 Batch 395/1077 - Train Accuracy: 0.5115, Validation Accuracy: 0.5277, Loss: 0.9646
Epoch 0 Batch 400/1077 - Train Accuracy: 0.4684, Validation Accuracy: 0.5320, Loss: 0.9979
Epoch 0 Batch 405/1077 - Train Accuracy: 0.4568, Validation Accuracy: 0.5298, Loss: 1.0415
Epoch 0 Batch 410/1077 - Train Accuracy: 0.4470, Validation Accuracy: 0.5366, Loss: 1.0260
Epoch 0 Batch 415/1077 - Train Accuracy: 0.5015, Validation Accuracy: 0.5359, Loss: 0.9450
Epoch 0 Batch 420/1077 - Train Accuracy: 0.4910, Validation Accuracy: 0.5437, Loss: 0.9645
Epoch 0 Batch 425/1077 - Train Accuracy: 0.5446, Validation Accuracy: 0.5394, Loss: 0.9257
Epoch 0 Batch 430/1077 - Train Accuracy: 0.4973, Validation Accuracy: 0.5469, Loss: 0.9471
Epoch 0 Batch 435/1077 - Train Accuracy: 0.4831, Validation Accuracy: 0.5295, Loss: 0.9968
Epoch 0 Batch 440/1077 - Train Accuracy: 0.5207, Validation Accuracy: 0.5518, Loss: 0.9646
Epoch 0 Batch 445/1077 - Train Accuracy: 0.4848, Validation Accuracy: 0.5408, Loss: 0.9771
Epoch 0 Batch 450/1077 - Train Accuracy: 0.4762, Validation Accuracy: 0.5348, Loss: 0.9330
Epoch 0 Batch 455/1077 - Train Accuracy: 0.5298, Validation Accuracy: 0.5504, Loss: 0.8659
Epoch 0 Batch 460/1077 - Train Accuracy: 0.4785, Validation Accuracy: 0.5447, Loss: 0.9152
Epoch 0 Batch 465/1077 - Train Accuracy: 0.4634, Validation Accuracy: 0.5415, Loss: 0.9382
Epoch 0 Batch 470/1077 - Train Accuracy: 0.4683, Validation Accuracy: 0.5387, Loss: 0.9340
Epoch 0 Batch 475/1077 - Train Accuracy: 0.5406, Validation Accuracy: 0.5540, Loss: 0.8482
Epoch 0 Batch 480/1077 - Train Accuracy: 0.4922, Validation Accuracy: 0.5508, Loss: 0.8971
Epoch 0 Batch 485/1077 - Train Accuracy: 0.5547, Validation Accuracy: 0.5490, Loss: 0.8493
Epoch 0 Batch 490/1077 - Train Accuracy: 0.5102, Validation Accuracy: 0.5476, Loss: 0.8771
Epoch 0 Batch 495/1077 - Train Accuracy: 0.5129, Validation Accuracy: 0.5483, Loss: 0.8648
Epoch 0 Batch 500/1077 - Train Accuracy: 0.5238, Validation Accuracy: 0.5451, Loss: 0.8492
Epoch 0 Batch 505/1077 - Train Accuracy: 0.5666, Validation Accuracy: 0.5508, Loss: 0.7669
Epoch 0 Batch 510/1077 - Train Accuracy: 0.5563, Validation Accuracy: 0.5522, Loss: 0.8442
Epoch 0 Batch 515/1077 - Train Accuracy: 0.5289, Validation Accuracy: 0.5543, Loss: 0.8565
Epoch 0 Batch 520/1077 - Train Accuracy: 0.5316, Validation Accuracy: 0.5547, Loss: 0.8016
Epoch 0 Batch 525/1077 - Train Accuracy: 0.5121, Validation Accuracy: 0.5582, Loss: 0.8266
Epoch 0 Batch 530/1077 - Train Accuracy: 0.5242, Validation Accuracy: 0.5600, Loss: 0.8288
Epoch 0 Batch 535/1077 - Train Accuracy: 0.5199, Validation Accuracy: 0.5604, Loss: 0.8232
Epoch 0 Batch 540/1077 - Train Accuracy: 0.5320, Validation Accuracy: 0.5586, Loss: 0.7885
Epoch 0 Batch 545/1077 - Train Accuracy: 0.5289, Validation Accuracy: 0.5607, Loss: 0.8512
Epoch 0 Batch 550/1077 - Train Accuracy: 0.5199, Validation Accuracy: 0.5536, Loss: 0.8394
Epoch 0 Batch 555/1077 - Train Accuracy: 0.5359, Validation Accuracy: 0.5568, Loss: 0.8136
Epoch 0 Batch 560/1077 - Train Accuracy: 0.5402, Validation Accuracy: 0.5604, Loss: 0.7884
Epoch 0 Batch 565/1077 - Train Accuracy: 0.5711, Validation Accuracy: 0.5732, Loss: 0.7767
Epoch 0 Batch 570/1077 - Train Accuracy: 0.5317, Validation Accuracy: 0.5703, Loss: 0.8306
Epoch 0 Batch 575/1077 - Train Accuracy: 0.5406, Validation Accuracy: 0.5827, Loss: 0.7810
Epoch 0 Batch 580/1077 - Train Accuracy: 0.5938, Validation Accuracy: 0.5756, Loss: 0.7210
Epoch 0 Batch 585/1077 - Train Accuracy: 0.5848, Validation Accuracy: 0.5795, Loss: 0.7176
Epoch 0 Batch 590/1077 - Train Accuracy: 0.5206, Validation Accuracy: 0.5717, Loss: 0.8007
Epoch 0 Batch 595/1077 - Train Accuracy: 0.5414, Validation Accuracy: 0.5749, Loss: 0.7814
Epoch 0 Batch 600/1077 - Train Accuracy: 0.5751, Validation Accuracy: 0.5742, Loss: 0.7263
Epoch 0 Batch 605/1077 - Train Accuracy: 0.5177, Validation Accuracy: 0.5788, Loss: 0.8328
Epoch 0 Batch 610/1077 - Train Accuracy: 0.5518, Validation Accuracy: 0.5795, Loss: 0.8061
Epoch 0 Batch 615/1077 - Train Accuracy: 0.5586, Validation Accuracy: 0.5803, Loss: 0.7556
Epoch 0 Batch 620/1077 - Train Accuracy: 0.5473, Validation Accuracy: 0.5838, Loss: 0.7495
Epoch 0 Batch 625/1077 - Train Accuracy: 0.5898, Validation Accuracy: 0.5827, Loss: 0.7442
Epoch 0 Batch 630/1077 - Train Accuracy: 0.5559, Validation Accuracy: 0.5803, Loss: 0.7417
Epoch 0 Batch 635/1077 - Train Accuracy: 0.5362, Validation Accuracy: 0.5810, Loss: 0.8056
Epoch 0 Batch 640/1077 - Train Accuracy: 0.5335, Validation Accuracy: 0.5820, Loss: 0.7234
Epoch 0 Batch 645/1077 - Train Accuracy: 0.5934, Validation Accuracy: 0.5842, Loss: 0.7048
Epoch 0 Batch 650/1077 - Train Accuracy: 0.5277, Validation Accuracy: 0.5870, Loss: 0.7597
Epoch 0 Batch 655/1077 - Train Accuracy: 0.5738, Validation Accuracy: 0.5881, Loss: 0.7522
Epoch 0 Batch 660/1077 - Train Accuracy: 0.5500, Validation Accuracy: 0.5902, Loss: 0.7522
Epoch 0 Batch 665/1077 - Train Accuracy: 0.5586, Validation Accuracy: 0.5831, Loss: 0.7223
Epoch 0 Batch 670/1077 - Train Accuracy: 0.5973, Validation Accuracy: 0.5820, Loss: 0.6779
Epoch 0 Batch 675/1077 - Train Accuracy: 0.5807, Validation Accuracy: 0.5845, Loss: 0.7108
Epoch 0 Batch 680/1077 - Train Accuracy: 0.5781, Validation Accuracy: 0.5845, Loss: 0.6970
Epoch 0 Batch 685/1077 - Train Accuracy: 0.5734, Validation Accuracy: 0.5785, Loss: 0.7288
Epoch 0 Batch 690/1077 - Train Accuracy: 0.5805, Validation Accuracy: 0.5813, Loss: 0.7104
Epoch 0 Batch 695/1077 - Train Accuracy: 0.5914, Validation Accuracy: 0.5952, Loss: 0.6910
Epoch 0 Batch 700/1077 - Train Accuracy: 0.5504, Validation Accuracy: 0.5881, Loss: 0.6965
Epoch 0 Batch 705/1077 - Train Accuracy: 0.5604, Validation Accuracy: 0.5941, Loss: 0.7670
Epoch 0 Batch 710/1077 - Train Accuracy: 0.5535, Validation Accuracy: 0.5881, Loss: 0.7017
Epoch 0 Batch 715/1077 - Train Accuracy: 0.5902, Validation Accuracy: 0.5934, Loss: 0.7112
Epoch 0 Batch 720/1077 - Train Accuracy: 0.5637, Validation Accuracy: 0.5763, Loss: 0.7357
Epoch 0 Batch 725/1077 - Train Accuracy: 0.5539, Validation Accuracy: 0.5916, Loss: 0.6535
Epoch 0 Batch 730/1077 - Train Accuracy: 0.5473, Validation Accuracy: 0.5906, Loss: 0.6991
Epoch 0 Batch 735/1077 - Train Accuracy: 0.5934, Validation Accuracy: 0.5945, Loss: 0.6747
Epoch 0 Batch 740/1077 - Train Accuracy: 0.5855, Validation Accuracy: 0.5906, Loss: 0.6733
Epoch 0 Batch 745/1077 - Train Accuracy: 0.5727, Validation Accuracy: 0.5831, Loss: 0.6784
Epoch 0 Batch 750/1077 - Train Accuracy: 0.5859, Validation Accuracy: 0.5987, Loss: 0.6757
Epoch 0 Batch 755/1077 - Train Accuracy: 0.5668, Validation Accuracy: 0.5952, Loss: 0.6835
Epoch 0 Batch 760/1077 - Train Accuracy: 0.6105, Validation Accuracy: 0.5998, Loss: 0.6823
Epoch 0 Batch 765/1077 - Train Accuracy: 0.5742, Validation Accuracy: 0.5824, Loss: 0.6438
Epoch 0 Batch 770/1077 - Train Accuracy: 0.5911, Validation Accuracy: 0.6040, Loss: 0.6375
Epoch 0 Batch 775/1077 - Train Accuracy: 0.6020, Validation Accuracy: 0.6009, Loss: 0.6641
Epoch 0 Batch 780/1077 - Train Accuracy: 0.5813, Validation Accuracy: 0.6037, Loss: 0.6857
Epoch 0 Batch 785/1077 - Train Accuracy: 0.6127, Validation Accuracy: 0.6005, Loss: 0.6280
Epoch 0 Batch 790/1077 - Train Accuracy: 0.5207, Validation Accuracy: 0.6019, Loss: 0.6873
Epoch 0 Batch 795/1077 - Train Accuracy: 0.5754, Validation Accuracy: 0.5923, Loss: 0.6784
Epoch 0 Batch 800/1077 - Train Accuracy: 0.5422, Validation Accuracy: 0.6080, Loss: 0.6445
Epoch 0 Batch 805/1077 - Train Accuracy: 0.6020, Validation Accuracy: 0.6136, Loss: 0.6446
Epoch 0 Batch 810/1077 - Train Accuracy: 0.6068, Validation Accuracy: 0.6030, Loss: 0.6128
Epoch 0 Batch 815/1077 - Train Accuracy: 0.5734, Validation Accuracy: 0.5987, Loss: 0.6453
Epoch 0 Batch 820/1077 - Train Accuracy: 0.5773, Validation Accuracy: 0.5998, Loss: 0.6690
Epoch 0 Batch 825/1077 - Train Accuracy: 0.5809, Validation Accuracy: 0.6143, Loss: 0.6262
Epoch 0 Batch 830/1077 - Train Accuracy: 0.5820, Validation Accuracy: 0.6158, Loss: 0.6406
Epoch 0 Batch 835/1077 - Train Accuracy: 0.5953, Validation Accuracy: 0.5966, Loss: 0.6416
Epoch 0 Batch 840/1077 - Train Accuracy: 0.6039, Validation Accuracy: 0.6143, Loss: 0.6224
Epoch 0 Batch 845/1077 - Train Accuracy: 0.5914, Validation Accuracy: 0.6087, Loss: 0.6269
Epoch 0 Batch 850/1077 - Train Accuracy: 0.5681, Validation Accuracy: 0.6040, Loss: 0.6682
Epoch 0 Batch 855/1077 - Train Accuracy: 0.5539, Validation Accuracy: 0.6037, Loss: 0.6337
Epoch 0 Batch 860/1077 - Train Accuracy: 0.5900, Validation Accuracy: 0.6090, Loss: 0.6241
Epoch 0 Batch 865/1077 - Train Accuracy: 0.6555, Validation Accuracy: 0.6168, Loss: 0.5708
Epoch 0 Batch 870/1077 - Train Accuracy: 0.5588, Validation Accuracy: 0.6062, Loss: 0.6795
Epoch 0 Batch 875/1077 - Train Accuracy: 0.5977, Validation Accuracy: 0.6033, Loss: 0.6428
Epoch 0 Batch 880/1077 - Train Accuracy: 0.6133, Validation Accuracy: 0.6005, Loss: 0.6240
Epoch 0 Batch 885/1077 - Train Accuracy: 0.6186, Validation Accuracy: 0.5920, Loss: 0.5451
Epoch 0 Batch 890/1077 - Train Accuracy: 0.6696, Validation Accuracy: 0.6033, Loss: 0.5961
Epoch 0 Batch 895/1077 - Train Accuracy: 0.5836, Validation Accuracy: 0.5987, Loss: 0.6172
Epoch 0 Batch 900/1077 - Train Accuracy: 0.6191, Validation Accuracy: 0.6055, Loss: 0.6275
Epoch 0 Batch 905/1077 - Train Accuracy: 0.6316, Validation Accuracy: 0.6083, Loss: 0.6014
Epoch 0 Batch 910/1077 - Train Accuracy: 0.5804, Validation Accuracy: 0.6051, Loss: 0.6173
Epoch 0 Batch 915/1077 - Train Accuracy: 0.5728, Validation Accuracy: 0.6101, Loss: 0.6454
Epoch 0 Batch 920/1077 - Train Accuracy: 0.5914, Validation Accuracy: 0.6229, Loss: 0.6226
Epoch 0 Batch 925/1077 - Train Accuracy: 0.5997, Validation Accuracy: 0.6250, Loss: 0.5980
Epoch 0 Batch 930/1077 - Train Accuracy: 0.6168, Validation Accuracy: 0.6104, Loss: 0.6036
Epoch 0 Batch 935/1077 - Train Accuracy: 0.6098, Validation Accuracy: 0.6101, Loss: 0.6258
Epoch 0 Batch 940/1077 - Train Accuracy: 0.5738, Validation Accuracy: 0.6119, Loss: 0.6155
Epoch 0 Batch 945/1077 - Train Accuracy: 0.6520, Validation Accuracy: 0.6232, Loss: 0.5851
Epoch 0 Batch 950/1077 - Train Accuracy: 0.5893, Validation Accuracy: 0.6300, Loss: 0.5792
Epoch 0 Batch 955/1077 - Train Accuracy: 0.6223, Validation Accuracy: 0.6179, Loss: 0.6139
Epoch 0 Batch 960/1077 - Train Accuracy: 0.6250, Validation Accuracy: 0.6168, Loss: 0.5776
Epoch 0 Batch 965/1077 - Train Accuracy: 0.5312, Validation Accuracy: 0.6115, Loss: 0.6411
Epoch 0 Batch 970/1077 - Train Accuracy: 0.6383, Validation Accuracy: 0.6083, Loss: 0.6053
Epoch 0 Batch 975/1077 - Train Accuracy: 0.6410, Validation Accuracy: 0.6204, Loss: 0.5618
Epoch 0 Batch 980/1077 - Train Accuracy: 0.6121, Validation Accuracy: 0.6168, Loss: 0.6017
Epoch 0 Batch 985/1077 - Train Accuracy: 0.6398, Validation Accuracy: 0.6300, Loss: 0.5913
Epoch 0 Batch 990/1077 - Train Accuracy: 0.6114, Validation Accuracy: 0.6424, Loss: 0.6343
Epoch 0 Batch 995/1077 - Train Accuracy: 0.6451, Validation Accuracy: 0.6293, Loss: 0.5774
Epoch 0 Batch 1000/1077 - Train Accuracy: 0.6384, Validation Accuracy: 0.6200, Loss: 0.5509
Epoch 0 Batch 1005/1077 - Train Accuracy: 0.5863, Validation Accuracy: 0.6232, Loss: 0.5954
Epoch 0 Batch 1010/1077 - Train Accuracy: 0.6141, Validation Accuracy: 0.6218, Loss: 0.5909
Epoch 0 Batch 1015/1077 - Train Accuracy: 0.5906, Validation Accuracy: 0.6218, Loss: 0.6251
Epoch 0 Batch 1020/1077 - Train Accuracy: 0.5809, Validation Accuracy: 0.6158, Loss: 0.5702
Epoch 0 Batch 1025/1077 - Train Accuracy: 0.6112, Validation Accuracy: 0.6207, Loss: 0.5632
Epoch 0 Batch 1030/1077 - Train Accuracy: 0.5926, Validation Accuracy: 0.6296, Loss: 0.5819
Epoch 0 Batch 1035/1077 - Train Accuracy: 0.6161, Validation Accuracy: 0.6310, Loss: 0.5380
Epoch 0 Batch 1040/1077 - Train Accuracy: 0.5863, Validation Accuracy: 0.6179, Loss: 0.6239
Epoch 0 Batch 1045/1077 - Train Accuracy: 0.5766, Validation Accuracy: 0.6183, Loss: 0.6033
Epoch 0 Batch 1050/1077 - Train Accuracy: 0.5496, Validation Accuracy: 0.6158, Loss: 0.5998
Epoch 0 Batch 1055/1077 - Train Accuracy: 0.6477, Validation Accuracy: 0.6218, Loss: 0.5815
Epoch 0 Batch 1060/1077 - Train Accuracy: 0.6156, Validation Accuracy: 0.6200, Loss: 0.5645
Epoch 0 Batch 1065/1077 - Train Accuracy: 0.6109, Validation Accuracy: 0.6200, Loss: 0.5749
Epoch 0 Batch 1070/1077 - Train Accuracy: 0.5746, Validation Accuracy: 0.6307, Loss: 0.5962
Epoch 0 Batch 1075/1077 - Train Accuracy: 0.5822, Validation Accuracy: 0.6076, Loss: 0.6106
Epoch 1 Batch 5/1077 - Train Accuracy: 0.5996, Validation Accuracy: 0.6175, Loss: 0.5985
Epoch 1 Batch 10/1077 - Train Accuracy: 0.6032, Validation Accuracy: 0.6264, Loss: 0.6020
Epoch 1 Batch 15/1077 - Train Accuracy: 0.6258, Validation Accuracy: 0.6261, Loss: 0.5724
Epoch 1 Batch 20/1077 - Train Accuracy: 0.6027, Validation Accuracy: 0.6332, Loss: 0.5607
Epoch 1 Batch 25/1077 - Train Accuracy: 0.6117, Validation Accuracy: 0.6261, Loss: 0.5491
Epoch 1 Batch 30/1077 - Train Accuracy: 0.6246, Validation Accuracy: 0.6300, Loss: 0.5745
Epoch 1 Batch 35/1077 - Train Accuracy: 0.6359, Validation Accuracy: 0.6371, Loss: 0.5738
Epoch 1 Batch 40/1077 - Train Accuracy: 0.6027, Validation Accuracy: 0.6310, Loss: 0.5788
Epoch 1 Batch 45/1077 - Train Accuracy: 0.6199, Validation Accuracy: 0.6300, Loss: 0.5803
Epoch 1 Batch 50/1077 - Train Accuracy: 0.5781, Validation Accuracy: 0.6449, Loss: 0.5687
Epoch 1 Batch 55/1077 - Train Accuracy: 0.6395, Validation Accuracy: 0.6353, Loss: 0.5657
Epoch 1 Batch 60/1077 - Train Accuracy: 0.6220, Validation Accuracy: 0.6257, Loss: 0.5495
Epoch 1 Batch 65/1077 - Train Accuracy: 0.5991, Validation Accuracy: 0.6222, Loss: 0.5902
Epoch 1 Batch 70/1077 - Train Accuracy: 0.5987, Validation Accuracy: 0.6392, Loss: 0.5856
Epoch 1 Batch 75/1077 - Train Accuracy: 0.6145, Validation Accuracy: 0.6222, Loss: 0.5772
Epoch 1 Batch 80/1077 - Train Accuracy: 0.6180, Validation Accuracy: 0.6346, Loss: 0.5648
Epoch 1 Batch 85/1077 - Train Accuracy: 0.6316, Validation Accuracy: 0.6346, Loss: 0.5538
Epoch 1 Batch 90/1077 - Train Accuracy: 0.6008, Validation Accuracy: 0.6271, Loss: 0.5740
Epoch 1 Batch 95/1077 - Train Accuracy: 0.6354, Validation Accuracy: 0.6431, Loss: 0.5806
Epoch 1 Batch 100/1077 - Train Accuracy: 0.6320, Validation Accuracy: 0.6353, Loss: 0.5637
Epoch 1 Batch 105/1077 - Train Accuracy: 0.6344, Validation Accuracy: 0.6364, Loss: 0.5667
Epoch 1 Batch 110/1077 - Train Accuracy: 0.6543, Validation Accuracy: 0.6388, Loss: 0.5296
Epoch 1 Batch 115/1077 - Train Accuracy: 0.6105, Validation Accuracy: 0.6286, Loss: 0.5762
Epoch 1 Batch 120/1077 - Train Accuracy: 0.6215, Validation Accuracy: 0.6410, Loss: 0.5823
Epoch 1 Batch 125/1077 - Train Accuracy: 0.6522, Validation Accuracy: 0.6374, Loss: 0.5353
Epoch 1 Batch 130/1077 - Train Accuracy: 0.6064, Validation Accuracy: 0.6204, Loss: 0.5278
Epoch 1 Batch 135/1077 - Train Accuracy: 0.6098, Validation Accuracy: 0.6193, Loss: 0.5752
Epoch 1 Batch 140/1077 - Train Accuracy: 0.5789, Validation Accuracy: 0.6186, Loss: 0.5785
Epoch 1 Batch 145/1077 - Train Accuracy: 0.6492, Validation Accuracy: 0.6232, Loss: 0.5743
Epoch 1 Batch 150/1077 - Train Accuracy: 0.6425, Validation Accuracy: 0.6310, Loss: 0.5267
Epoch 1 Batch 155/1077 - Train Accuracy: 0.5949, Validation Accuracy: 0.6275, Loss: 0.5566
Epoch 1 Batch 160/1077 - Train Accuracy: 0.6375, Validation Accuracy: 0.6254, Loss: 0.5456
Epoch 1 Batch 165/1077 - Train Accuracy: 0.6008, Validation Accuracy: 0.6303, Loss: 0.5546
Epoch 1 Batch 170/1077 - Train Accuracy: 0.5820, Validation Accuracy: 0.6328, Loss: 0.5766
Epoch 1 Batch 175/1077 - Train Accuracy: 0.6227, Validation Accuracy: 0.6293, Loss: 0.5383
Epoch 1 Batch 180/1077 - Train Accuracy: 0.6141, Validation Accuracy: 0.6271, Loss: 0.5561
Epoch 1 Batch 185/1077 - Train Accuracy: 0.5871, Validation Accuracy: 0.6289, Loss: 0.5799
Epoch 1 Batch 190/1077 - Train Accuracy: 0.6602, Validation Accuracy: 0.6374, Loss: 0.5266
Epoch 1 Batch 195/1077 - Train Accuracy: 0.5848, Validation Accuracy: 0.6232, Loss: 0.5565
Epoch 1 Batch 200/1077 - Train Accuracy: 0.5887, Validation Accuracy: 0.6229, Loss: 0.5677
Epoch 1 Batch 205/1077 - Train Accuracy: 0.6316, Validation Accuracy: 0.6268, Loss: 0.5407
Epoch 1 Batch 210/1077 - Train Accuracy: 0.6417, Validation Accuracy: 0.6193, Loss: 0.5307
Epoch 1 Batch 215/1077 - Train Accuracy: 0.6016, Validation Accuracy: 0.6151, Loss: 0.5620
Epoch 1 Batch 220/1077 - Train Accuracy: 0.6201, Validation Accuracy: 0.6374, Loss: 0.5532
Epoch 1 Batch 225/1077 - Train Accuracy: 0.6168, Validation Accuracy: 0.6378, Loss: 0.5630
Epoch 1 Batch 230/1077 - Train Accuracy: 0.6484, Validation Accuracy: 0.6303, Loss: 0.5334
Epoch 1 Batch 235/1077 - Train Accuracy: 0.6693, Validation Accuracy: 0.6243, Loss: 0.5004
Epoch 1 Batch 240/1077 - Train Accuracy: 0.6324, Validation Accuracy: 0.6158, Loss: 0.5301
Epoch 1 Batch 245/1077 - Train Accuracy: 0.6224, Validation Accuracy: 0.6314, Loss: 0.5168
Epoch 1 Batch 250/1077 - Train Accuracy: 0.5941, Validation Accuracy: 0.6317, Loss: 0.5089
Epoch 1 Batch 255/1077 - Train Accuracy: 0.6117, Validation Accuracy: 0.6236, Loss: 0.5360
Epoch 1 Batch 260/1077 - Train Accuracy: 0.6291, Validation Accuracy: 0.6232, Loss: 0.5047
Epoch 1 Batch 265/1077 - Train Accuracy: 0.5996, Validation Accuracy: 0.6261, Loss: 0.5525
Epoch 1 Batch 270/1077 - Train Accuracy: 0.5949, Validation Accuracy: 0.6300, Loss: 0.5549
Epoch 1 Batch 275/1077 - Train Accuracy: 0.6086, Validation Accuracy: 0.6126, Loss: 0.5160
Epoch 1 Batch 280/1077 - Train Accuracy: 0.6090, Validation Accuracy: 0.6250, Loss: 0.5528
Epoch 1 Batch 285/1077 - Train Accuracy: 0.6399, Validation Accuracy: 0.6250, Loss: 0.5180
Epoch 1 Batch 290/1077 - Train Accuracy: 0.6055, Validation Accuracy: 0.6261, Loss: 0.5638
Epoch 1 Batch 295/1077 - Train Accuracy: 0.6456, Validation Accuracy: 0.6321, Loss: 0.5648
Epoch 1 Batch 300/1077 - Train Accuracy: 0.6205, Validation Accuracy: 0.6332, Loss: 0.5409
Epoch 1 Batch 305/1077 - Train Accuracy: 0.6473, Validation Accuracy: 0.6307, Loss: 0.5153
Epoch 1 Batch 310/1077 - Train Accuracy: 0.5988, Validation Accuracy: 0.6204, Loss: 0.5532
Epoch 1 Batch 315/1077 - Train Accuracy: 0.6343, Validation Accuracy: 0.6214, Loss: 0.4952
Epoch 1 Batch 320/1077 - Train Accuracy: 0.6438, Validation Accuracy: 0.6435, Loss: 0.5247
Epoch 1 Batch 325/1077 - Train Accuracy: 0.6589, Validation Accuracy: 0.6438, Loss: 0.5153
Epoch 1 Batch 330/1077 - Train Accuracy: 0.6469, Validation Accuracy: 0.6193, Loss: 0.5157
Epoch 1 Batch 335/1077 - Train Accuracy: 0.6466, Validation Accuracy: 0.6282, Loss: 0.5103
Epoch 1 Batch 340/1077 - Train Accuracy: 0.5876, Validation Accuracy: 0.6197, Loss: 0.5442
Epoch 1 Batch 345/1077 - Train Accuracy: 0.6559, Validation Accuracy: 0.6243, Loss: 0.5144
Epoch 1 Batch 350/1077 - Train Accuracy: 0.6008, Validation Accuracy: 0.6239, Loss: 0.5370
Epoch 1 Batch 355/1077 - Train Accuracy: 0.6161, Validation Accuracy: 0.6225, Loss: 0.5102
Epoch 1 Batch 360/1077 - Train Accuracy: 0.6191, Validation Accuracy: 0.6371, Loss: 0.5161
Epoch 1 Batch 365/1077 - Train Accuracy: 0.5965, Validation Accuracy: 0.6321, Loss: 0.5296
Epoch 1 Batch 370/1077 - Train Accuracy: 0.6376, Validation Accuracy: 0.6335, Loss: 0.5137
Epoch 1 Batch 375/1077 - Train Accuracy: 0.6598, Validation Accuracy: 0.6491, Loss: 0.4865
Epoch 1 Batch 380/1077 - Train Accuracy: 0.6426, Validation Accuracy: 0.6367, Loss: 0.5186
Epoch 1 Batch 385/1077 - Train Accuracy: 0.6312, Validation Accuracy: 0.6381, Loss: 0.5282
Epoch 1 Batch 390/1077 - Train Accuracy: 0.5754, Validation Accuracy: 0.6317, Loss: 0.5454
Epoch 1 Batch 395/1077 - Train Accuracy: 0.6518, Validation Accuracy: 0.6360, Loss: 0.4965
Epoch 1 Batch 400/1077 - Train Accuracy: 0.6645, Validation Accuracy: 0.6410, Loss: 0.5279
Epoch 1 Batch 405/1077 - Train Accuracy: 0.6209, Validation Accuracy: 0.6385, Loss: 0.5774
Epoch 1 Batch 410/1077 - Train Accuracy: 0.6221, Validation Accuracy: 0.6477, Loss: 0.5409
Epoch 1 Batch 415/1077 - Train Accuracy: 0.6477, Validation Accuracy: 0.6420, Loss: 0.4855
Epoch 1 Batch 420/1077 - Train Accuracy: 0.6020, Validation Accuracy: 0.6303, Loss: 0.5089
Epoch 1 Batch 425/1077 - Train Accuracy: 0.6510, Validation Accuracy: 0.6183, Loss: 0.5038
Epoch 1 Batch 430/1077 - Train Accuracy: 0.6059, Validation Accuracy: 0.6175, Loss: 0.5206
Epoch 1 Batch 435/1077 - Train Accuracy: 0.6353, Validation Accuracy: 0.6488, Loss: 0.5409
Epoch 1 Batch 440/1077 - Train Accuracy: 0.6238, Validation Accuracy: 0.6346, Loss: 0.5324
Epoch 1 Batch 445/1077 - Train Accuracy: 0.5888, Validation Accuracy: 0.6264, Loss: 0.5373
Epoch 1 Batch 450/1077 - Train Accuracy: 0.5879, Validation Accuracy: 0.6371, Loss: 0.4933
Epoch 1 Batch 455/1077 - Train Accuracy: 0.6346, Validation Accuracy: 0.6214, Loss: 0.4949
Epoch 1 Batch 460/1077 - Train Accuracy: 0.6387, Validation Accuracy: 0.6335, Loss: 0.5275
Epoch 1 Batch 465/1077 - Train Accuracy: 0.6414, Validation Accuracy: 0.6332, Loss: 0.5563
Epoch 1 Batch 470/1077 - Train Accuracy: 0.5958, Validation Accuracy: 0.6325, Loss: 0.5396
Epoch 1 Batch 475/1077 - Train Accuracy: 0.6395, Validation Accuracy: 0.6286, Loss: 0.5121
Epoch 1 Batch 480/1077 - Train Accuracy: 0.6517, Validation Accuracy: 0.6360, Loss: 0.5179
Epoch 1 Batch 485/1077 - Train Accuracy: 0.6656, Validation Accuracy: 0.6353, Loss: 0.5270
Epoch 1 Batch 490/1077 - Train Accuracy: 0.6117, Validation Accuracy: 0.6357, Loss: 0.5264
Epoch 1 Batch 495/1077 - Train Accuracy: 0.6234, Validation Accuracy: 0.6317, Loss: 0.4959
Epoch 1 Batch 500/1077 - Train Accuracy: 0.6258, Validation Accuracy: 0.6495, Loss: 0.5059
Epoch 1 Batch 505/1077 - Train Accuracy: 0.6611, Validation Accuracy: 0.6463, Loss: 0.4723
Epoch 1 Batch 510/1077 - Train Accuracy: 0.6305, Validation Accuracy: 0.6388, Loss: 0.5007
Epoch 1 Batch 515/1077 - Train Accuracy: 0.6191, Validation Accuracy: 0.6335, Loss: 0.5261
Epoch 1 Batch 520/1077 - Train Accuracy: 0.6656, Validation Accuracy: 0.6449, Loss: 0.4775
Epoch 1 Batch 525/1077 - Train Accuracy: 0.6438, Validation Accuracy: 0.6445, Loss: 0.5058
Epoch 1 Batch 530/1077 - Train Accuracy: 0.6219, Validation Accuracy: 0.6534, Loss: 0.5169
Epoch 1 Batch 535/1077 - Train Accuracy: 0.6129, Validation Accuracy: 0.6374, Loss: 0.5123
Epoch 1 Batch 540/1077 - Train Accuracy: 0.6422, Validation Accuracy: 0.6424, Loss: 0.4728
Epoch 1 Batch 545/1077 - Train Accuracy: 0.6348, Validation Accuracy: 0.6420, Loss: 0.5339
Epoch 1 Batch 550/1077 - Train Accuracy: 0.6172, Validation Accuracy: 0.6413, Loss: 0.5250
Epoch 1 Batch 555/1077 - Train Accuracy: 0.6484, Validation Accuracy: 0.6357, Loss: 0.5100
Epoch 1 Batch 560/1077 - Train Accuracy: 0.6129, Validation Accuracy: 0.6349, Loss: 0.4940
Epoch 1 Batch 565/1077 - Train Accuracy: 0.6414, Validation Accuracy: 0.6417, Loss: 0.4974
Epoch 1 Batch 570/1077 - Train Accuracy: 0.6386, Validation Accuracy: 0.6399, Loss: 0.5339
Epoch 1 Batch 575/1077 - Train Accuracy: 0.6410, Validation Accuracy: 0.6278, Loss: 0.4950
Epoch 1 Batch 580/1077 - Train Accuracy: 0.6775, Validation Accuracy: 0.6484, Loss: 0.4654
Epoch 1 Batch 585/1077 - Train Accuracy: 0.6722, Validation Accuracy: 0.6442, Loss: 0.4788
Epoch 1 Batch 590/1077 - Train Accuracy: 0.6172, Validation Accuracy: 0.6438, Loss: 0.5276
Epoch 1 Batch 595/1077 - Train Accuracy: 0.6125, Validation Accuracy: 0.6278, Loss: 0.4971
Epoch 1 Batch 600/1077 - Train Accuracy: 0.6678, Validation Accuracy: 0.6388, Loss: 0.4639
Epoch 1 Batch 605/1077 - Train Accuracy: 0.6760, Validation Accuracy: 0.6499, Loss: 0.5144
Epoch 1 Batch 610/1077 - Train Accuracy: 0.6250, Validation Accuracy: 0.6477, Loss: 0.5105
Epoch 1 Batch 615/1077 - Train Accuracy: 0.6570, Validation Accuracy: 0.6484, Loss: 0.4885
Epoch 1 Batch 620/1077 - Train Accuracy: 0.6324, Validation Accuracy: 0.6488, Loss: 0.4772
Epoch 1 Batch 625/1077 - Train Accuracy: 0.6777, Validation Accuracy: 0.6470, Loss: 0.4922
Epoch 1 Batch 630/1077 - Train Accuracy: 0.6418, Validation Accuracy: 0.6470, Loss: 0.4938
Epoch 1 Batch 635/1077 - Train Accuracy: 0.6225, Validation Accuracy: 0.6545, Loss: 0.5189
Epoch 1 Batch 640/1077 - Train Accuracy: 0.6298, Validation Accuracy: 0.6502, Loss: 0.4754
Epoch 1 Batch 645/1077 - Train Accuracy: 0.6633, Validation Accuracy: 0.6403, Loss: 0.4588
Epoch 1 Batch 650/1077 - Train Accuracy: 0.5848, Validation Accuracy: 0.6442, Loss: 0.5034
Epoch 1 Batch 655/1077 - Train Accuracy: 0.6285, Validation Accuracy: 0.6367, Loss: 0.5056
Epoch 1 Batch 660/1077 - Train Accuracy: 0.6398, Validation Accuracy: 0.6438, Loss: 0.5187
Epoch 1 Batch 665/1077 - Train Accuracy: 0.6062, Validation Accuracy: 0.6410, Loss: 0.5000
Epoch 1 Batch 670/1077 - Train Accuracy: 0.6697, Validation Accuracy: 0.6562, Loss: 0.4637
Epoch 1 Batch 675/1077 - Train Accuracy: 0.6499, Validation Accuracy: 0.6619, Loss: 0.4753
Epoch 1 Batch 680/1077 - Train Accuracy: 0.6298, Validation Accuracy: 0.6495, Loss: 0.4741
Epoch 1 Batch 685/1077 - Train Accuracy: 0.6281, Validation Accuracy: 0.6357, Loss: 0.4853
Epoch 1 Batch 690/1077 - Train Accuracy: 0.6430, Validation Accuracy: 0.6406, Loss: 0.4872
Epoch 1 Batch 695/1077 - Train Accuracy: 0.6566, Validation Accuracy: 0.6570, Loss: 0.4714
Epoch 1 Batch 700/1077 - Train Accuracy: 0.6020, Validation Accuracy: 0.6580, Loss: 0.4759
Epoch 1 Batch 705/1077 - Train Accuracy: 0.6682, Validation Accuracy: 0.6392, Loss: 0.5345
Epoch 1 Batch 710/1077 - Train Accuracy: 0.6070, Validation Accuracy: 0.6445, Loss: 0.4869
Epoch 1 Batch 715/1077 - Train Accuracy: 0.6465, Validation Accuracy: 0.6559, Loss: 0.4783
Epoch 1 Batch 720/1077 - Train Accuracy: 0.6562, Validation Accuracy: 0.6598, Loss: 0.5079
Epoch 1 Batch 725/1077 - Train Accuracy: 0.6481, Validation Accuracy: 0.6506, Loss: 0.4606
Epoch 1 Batch 730/1077 - Train Accuracy: 0.6254, Validation Accuracy: 0.6360, Loss: 0.4844
Epoch 1 Batch 735/1077 - Train Accuracy: 0.6242, Validation Accuracy: 0.6467, Loss: 0.4680
Epoch 1 Batch 740/1077 - Train Accuracy: 0.6520, Validation Accuracy: 0.6534, Loss: 0.4687
Epoch 1 Batch 745/1077 - Train Accuracy: 0.6727, Validation Accuracy: 0.6474, Loss: 0.4603
Epoch 1 Batch 750/1077 - Train Accuracy: 0.6570, Validation Accuracy: 0.6602, Loss: 0.4755
Epoch 1 Batch 755/1077 - Train Accuracy: 0.6430, Validation Accuracy: 0.6612, Loss: 0.4838
Epoch 1 Batch 760/1077 - Train Accuracy: 0.6516, Validation Accuracy: 0.6591, Loss: 0.4726
Epoch 1 Batch 765/1077 - Train Accuracy: 0.6641, Validation Accuracy: 0.6619, Loss: 0.4555
Epoch 1 Batch 770/1077 - Train Accuracy: 0.6380, Validation Accuracy: 0.6527, Loss: 0.4489
Epoch 1 Batch 775/1077 - Train Accuracy: 0.6480, Validation Accuracy: 0.6598, Loss: 0.4614
Epoch 1 Batch 780/1077 - Train Accuracy: 0.6570, Validation Accuracy: 0.6651, Loss: 0.4792
Epoch 1 Batch 785/1077 - Train Accuracy: 0.6775, Validation Accuracy: 0.6602, Loss: 0.4533
Epoch 1 Batch 790/1077 - Train Accuracy: 0.5867, Validation Accuracy: 0.6651, Loss: 0.4935
Epoch 1 Batch 795/1077 - Train Accuracy: 0.6531, Validation Accuracy: 0.6562, Loss: 0.4834
Epoch 1 Batch 800/1077 - Train Accuracy: 0.6035, Validation Accuracy: 0.6566, Loss: 0.4692
Epoch 1 Batch 805/1077 - Train Accuracy: 0.6805, Validation Accuracy: 0.6626, Loss: 0.4646
Epoch 1 Batch 810/1077 - Train Accuracy: 0.6559, Validation Accuracy: 0.6488, Loss: 0.4372
Epoch 1 Batch 815/1077 - Train Accuracy: 0.6281, Validation Accuracy: 0.6523, Loss: 0.4664
Epoch 1 Batch 820/1077 - Train Accuracy: 0.6387, Validation Accuracy: 0.6520, Loss: 0.4948
Epoch 1 Batch 825/1077 - Train Accuracy: 0.6312, Validation Accuracy: 0.6467, Loss: 0.4574
Epoch 1 Batch 830/1077 - Train Accuracy: 0.6441, Validation Accuracy: 0.6570, Loss: 0.4667
Epoch 1 Batch 835/1077 - Train Accuracy: 0.6711, Validation Accuracy: 0.6673, Loss: 0.4413
Epoch 1 Batch 840/1077 - Train Accuracy: 0.6527, Validation Accuracy: 0.6680, Loss: 0.4422
Epoch 1 Batch 845/1077 - Train Accuracy: 0.6500, Validation Accuracy: 0.6641, Loss: 0.4607
Epoch 1 Batch 850/1077 - Train Accuracy: 0.6202, Validation Accuracy: 0.6612, Loss: 0.4880
Epoch 1 Batch 855/1077 - Train Accuracy: 0.6348, Validation Accuracy: 0.6570, Loss: 0.4619
---------------------------------------------------------------------------
KeyboardInterrupt Traceback (most recent call last)
<ipython-input-50-2fda6f94fad5> in <module>()
46 target_sequence_length: targets_lengths,
47 source_sequence_length: sources_lengths,
---> 48 keep_prob: keep_probability})
49
50
d:\Anaconda3\envs\tf1.1\lib\site-packages\tensorflow\python\client\session.py in run(self, fetches, feed_dict, options, run_metadata)
776 try:
777 result = self._run(None, fetches, feed_dict, options_ptr,
--> 778 run_metadata_ptr)
779 if run_metadata:
780 proto_data = tf_session.TF_GetBuffer(run_metadata_ptr)
d:\Anaconda3\envs\tf1.1\lib\site-packages\tensorflow\python\client\session.py in _run(self, handle, fetches, feed_dict, options, run_metadata)
980 if final_fetches or final_targets:
981 results = self._do_run(handle, final_targets, final_fetches,
--> 982 feed_dict_string, options, run_metadata)
983 else:
984 results = []
d:\Anaconda3\envs\tf1.1\lib\site-packages\tensorflow\python\client\session.py in _do_run(self, handle, target_list, fetch_list, feed_dict, options, run_metadata)
1030 if handle is None:
1031 return self._do_call(_run_fn, self._session, feed_dict, fetch_list,
-> 1032 target_list, options, run_metadata)
1033 else:
1034 return self._do_call(_prun_fn, self._session, handle, feed_dict,
d:\Anaconda3\envs\tf1.1\lib\site-packages\tensorflow\python\client\session.py in _do_call(self, fn, *args)
1037 def _do_call(self, fn, *args):
1038 try:
-> 1039 return fn(*args)
1040 except errors.OpError as e:
1041 message = compat.as_text(e.message)
d:\Anaconda3\envs\tf1.1\lib\site-packages\tensorflow\python\client\session.py in _run_fn(session, feed_dict, fetch_list, target_list, options, run_metadata)
1019 return tf_session.TF_Run(session, options,
1020 feed_dict, fetch_list, target_list,
-> 1021 status, run_metadata)
1022
1023 def _prun_fn(session, handle, feed_dict, fetch_list):
KeyboardInterrupt:
保存参数
"""
DON'T MODIFY ANYTHING IN THIS CELL
"""
# Save parameters for checkpoint
helper.save_params(save_path)
"""
DON'T MODIFY ANYTHING IN THIS CELL
"""
import tensorflow as tf
import numpy as np
import helper
import problem_unittests as tests
_, (source_vocab_to_int, target_vocab_to_int), (source_int_to_vocab, target_int_to_vocab) = helper.load_preprocess()
load_path = helper.load_params()
句子转换为序列
def sentence_to_seq(sentence, vocab_to_int):
"""
Convert a sentence to a sequence of ids
:param sentence: String
:param vocab_to_int: Dictionary to go from the words to an id
:return: List of word ids
"""
# TODO: Implement Function
ids = []
for word in sentence.split():
idx = vocab_to_int.get(word, vocab_to_int['<UNK>'])
ids.append(idx)
return ids
"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
tests.test_sentence_to_seq(sentence_to_seq)
利用训练好的网络翻译一个英语句子。
translate_sentence = 'he saw a old yellow truck .'
"""
DON'T MODIFY ANYTHING IN THIS CELL
"""
translate_sentence = sentence_to_seq(translate_sentence, source_vocab_to_int)
loaded_graph = tf.Graph()
with tf.Session(graph=loaded_graph) as sess:
# Load saved model
loader = tf.train.import_meta_graph(load_path + '.meta')
loader.restore(sess, load_path)
input_data = loaded_graph.get_tensor_by_name('input:0')
logits = loaded_graph.get_tensor_by_name('predictions:0')
target_sequence_length = loaded_graph.get_tensor_by_name('target_sequence_length:0')
source_sequence_length = loaded_graph.get_tensor_by_name('source_sequence_length:0')
keep_prob = loaded_graph.get_tensor_by_name('keep_prob:0')
translate_logits = sess.run(logits, {input_data: [translate_sentence]*batch_size,
target_sequence_length: [len(translate_sentence)*2]*batch_size,
source_sequence_length: [len(translate_sentence)]*batch_size,
keep_prob: 1.0})[0]
print('Input')
print(' Word Ids: {}'.format([i for i in translate_sentence]))
print(' English Words: {}'.format([source_int_to_vocab[i] for i in translate_sentence]))
print('\nPrediction')
print(' Word Ids: {}'.format([i for i in translate_logits]))
print(' French Words: {}'.format(" ".join([target_int_to_vocab[i] for i in translate_logits])))
要想得到更好的结果，需要更多的训练数据集，你可以使用这个 WMT10 French-English corpus，但是这也将会花费更长的时间。