In [1]:

    

import numpy as np
import tensorflow as tf
import tempfile
tf.reset_default_graph()
sess = tf.InteractiveSession()

seq_length = 5
batch_size = 64

vocab_size = 7
embedding_dim = 50

memory_dim = 100


    

In [2]:

    

i=0
y_train=[]
for line in open('/Volumes/Zhipeng/patent_dataset/paired_newclaims_2012to2013.txt'):
    if i%2==0:
        y_train.append(line)
    i+=1


    

In [3]:

    

i=0
x_train=[]
for line in open('/Volumes/Zhipeng/patent_dataset/paired_oldclaims_2012to2013.txt'):
    if i%2==0:
        x_train.append(line)
    i+=1


    

In [4]:

    

len(x_train)


    

    
Out[4]:





155152

In [5]:

    

enc_inp = [tf.placeholder(tf.int32, shape=(None,),
                          name="inp%i" % t)
           for t in range(seq_length)]

labels = [tf.placeholder(tf.int32, shape=(None,),
                        name="labels%i" % t)
          for t in range(seq_length)]

weights = [tf.ones_like(labels_t, dtype=tf.float32)
           for labels_t in labels]


    

In [6]:

    

# Decoder input: prepend some "GO" token and drop the final
# token of the encoder input
dec_inp = ([tf.zeros_like(enc_inp[0], dtype=np.int32, name="GO")]
           + enc_inp[:-1])


    

In [7]:

    

# Initial memory value for recurrence.
prev_mem = tf.zeros((batch_size, memory_dim))

cell = tf.nn.rnn_cell.GRUCell(memory_dim)

dec_outputs, dec_memory = tf.nn.seq2seq.embedding_rnn_seq2seq(
    enc_inp, dec_inp, cell, vocab_size, vocab_size,embedding_size=embedding_dim)

loss = tf.nn.seq2seq.sequence_loss(dec_outputs, labels, weights, vocab_size)

tf.scalar_summary("loss", loss)

magnitude = tf.sqrt(tf.reduce_sum(tf.square(dec_memory[1])))
tf.scalar_summary("magnitude at t=1", magnitude)

summary_op = tf.merge_all_summaries()


    

In [8]:

    

learning_rate = 0.05
momentum = 0.9
optimizer = tf.train.MomentumOptimizer(learning_rate, momentum)
train_op = optimizer.minimize(loss)

logdir = tempfile.mkdtemp()
print (logdir)
summary_writer = tf.train.SummaryWriter(logdir, sess.graph_def)

sess.run(tf.initialize_all_variables())


    

    




/var/folders/br/8ftyb7yd2vn4mzhlz4tqcmn00000gn/T/tmptcwgjv05






    




WARNING:tensorflow:Passing a `GraphDef` to the SummaryWriter is deprecated. Pass a `Graph` object instead, such as `sess.graph`.

In [ ]:

    

def train_batch(batch_size):
    #X = [np.random.choice(vocab_size, size=(seq_length,), replace=False)
    #     for _ in range(batch_size)]
    #Y = X[:]
    # Dimshuffle to seq_len * batch_size
    X=x_train
    Y=y_train
    X = np.array(X).T
    Y = np.array(Y).T

    feed_dict = {enc_inp[t]: X[t] for t in range(seq_length)}
    feed_dict.update({labels[t]: Y[t] for t in range(seq_length)})

    output, loss_t, summary = sess.run([train_op, loss, summary_op], feed_dict)
    return output,loss_t, summary
    
for t in range(500):
    output,loss_t, summary = train_batch(batch_size)
    summary_writer.add_summary(summary, t)
summary_writer.flush()


    

In [ ]:

    

X_batch = [np.random.choice(vocab_size, size=(seq_length,), replace=False)
           for _ in range(10)]
X_batch = np.array(X_batch).T

feed_dict = {enc_inp[t]: X_batch[t] for t in range(seq_length)}
dec_outputs_batch = sess.run(dec_outputs, feed_dict)

#X_batch
#Y=[logits_t.argmax(axis=1) for logits_t in dec_outputs_batch]


    

In [ ]:

    




    

In [ ]: