Tutorial on word2vector using GloVe and Word2Vec
Tutorial on word2vector using GloVe and Word2Vec
2018-05-04 10:02:53
Some Important Reference Pages First:
Reference Page: https://github.com/IliaGavrilov/NeuralMachineTranslationBidirectionalLSTM/blob/master/1_Bidirectional_LSTM_Eng_to_French.ipynb
Glove Project Page: https://nlp.stanford.edu/projects/glove/
Word2Vec Project Page: https://code.google.com/archive/p/word2vec/
Pre-trained word2vec model: https://drive.google.com/file/d/0B7XkCwpI5KDYNlNUTTlSS21pQmM/edit?usp=sharing
Gensim Tutorial: https://radimrehurek.com/gensim/models/word2vec.html
===================================
===== For the Glove
===================================
1. Download one of the pre-trained model from Glove project page and Unzip the files.
1 Wikipedia 2014 + Gigaword 5 (6B tokens, 400K vocab, uncased, 50d, 100d, 200d, & 300d vectors, 822 MB download): glove.6B.zip
2 Common Crawl (42B tokens, 1.9M vocab, uncased, 300d vectors, 1.75 GB download): glove.42B.300d.zip
3 Common Crawl (840B tokens, 2.2M vocab, cased, 300d vectors, 2.03 GB download): glove.840B.300d.zip
2. Install the needed packages:
pickle, numpy, re, pickle, collections, bcolz
3. Run the following demo to test the results (extract the feature of given single word).
Code:
1 import pickle
2 import numpy as np
3 import re, pickle, collections, bcolz
4
5 # with open('./glove.840B.300d.txt', 'r', encoding="utf8") as f:
6
7 with open('./glove.6B.200d.txt', 'r') as f:
8 lines = [line.split() for line in f]
9
10 print('==>> begin to load Glove pre-trained models.')
11 glove_words = [elem[0] for elem in lines]
12 glove_words_idx = {elem:idx for idx, elem in enumerate(glove_words)} # is elem: idx equal to glove_words_idx[elem]=idx?
13 glove_vecs = np.stack(np.array(elem[1:], dtype=np.float32) for elem in lines)
14
15 print('==>> save into .pkl files.')
16 pickle.dump(glove_words, open('./glove.6B.200d.txt'+'_glove_words.pkl', 'wb'))
17 pickle.dump(glove_words_idx, open('./glove.6B.200d.txt'+'_glove_words_idx.pkl', 'wb'))
18
19 ## saving array using specific function.
20 def save_array(fname, arr):
21 c=bcolz.carray(arr, rootdir=fname, mode='w')
22 c.flush()
23
24 save_array('./glove.6B.200d.txt'+'_glove_vecs'+'.dat', glove_vecs)
25
26 def load_glove(loc):
27 return (bcolz.open(loc+'_glove_vecs.dat')[:],
28 pickle.load(open(loc+'_glove_words.pkl', 'rb')),
29 pickle.load(open(loc+'_glove_words_idx.pkl', 'rb')))
30
31
32 ###############################################
33 print('==>> Loading the glove.6B.200d.txt files.')
34 en_vecs, en_wv_word, en_wv_idx = load_glove('./glove.6B.200d.txt')
35 en_w2v = {w: en_vecs[en_wv_idx[w]] for w in en_wv_word}
36 n_en_vec, dim_en_vec = en_vecs.shape
37
38 print('==>> shown one demo: "King"')
39 demo_vector = en_w2v['king']
40 print(demo_vector)
41 print("==>> Done !")
Results:
wangxiao@AHU$ python tutorial_Glove_word2vec.py ==>> begin to load Glove pre-trained models. ==>> save into .pkl files. ==>> Loading the glove.6B.200d.txt files. ==>> shown one demo: "King" [-0.49346 -0.14768 0.32166001 0.056899 0.052572 0.20192 -0.13506 -0.030793 0.15614 -0.23004 -0.66376001 -0.27316001 0.10391 0.57334 -0.032355 -0.32765999 -0.27160001 0.32918999 0.41305 -0.18085 1.51670003 2.16490006 -0.10278 0.098019 -0.018946 0.027292 -0.79479998 0.36631 -0.33151001 0.28839999 0.10436 -0.19166 0.27326 -0.17519 -0.14985999 -0.072333 -0.54370999 -0.29728001 0.081491 -0.42673001 -0.36406001 -0.52034998 0.18455 0.44121 -0.32196 0.39172 0.11952 0.36978999 0.29229 -0.42954001 0.46653 -0.067243 0.31215999 -0.17216 0.48874 0.28029999 -0.17577 -0.35100999 0.020792 0.15974 0.21927001 -0.32499 0.086022 0.38927001 -0.65638 -0.67400998 -0.41896001 1.27090001 0.20857 0.28314999 0.58238 -0.14944001 0.3989 0.52680999 0.35714 -0.39100999 -0.55372 -0.56642002 -0.15762 -0.48004001 0.40448001 0.057518 -1.01569998 0.21754999 0.073296 0.15237001 -0.38361999 -0.75308001 -0.0060254 -0.26232001 -0.54101998 -0.34347001 0.11113 0.47685 -0.73229998 0.77596998 0.015216 -0.66327 -0.21144 -0.42964 -0.72689998 -0.067968 0.50601 0.039817 -0.27584001 -0.34794 -0.0474 0.50734001 -0.30777001 0.11594 -0.19211 0.3107 -0.60075003 0.22044 -0.36265001 -0.59442002 -1.20459998 0.10619 -0.60277998 0.21573 -0.35361999 0.55473 0.58094001 0.077259 1.0776 -0.1867 -1.51680005 0.32418001 0.83332998 0.17366 1.12320006 0.10863 0.55888999 0.30799001 0.084318 -0.43178001 -0.042287 -0.054615 0.054712 -0.80914003 -0.24429999 -0.076909 0.55216002 -0.71895999 0.83319002 0.020735 0.020472 -0.40279001 -0.28874001 0.23758 0.12576 -0.15165 -0.69419998 -0.25174001 0.29591 0.40290001 -1.0618 0.19847 -0.63463002 -0.70842999 0.067943 0.57366002 0.041122 0.17452 0.19430999 -0.28641 -1.13629997 0.45116001 -0.066518 0.82615 -0.45451999 -0.85652 0.18105 -0.24187 0.20152999 0.72298002 0.17415 -0.87327999 0.69814998 0.024706 0.26174 -0.0087155 -0.39348999 0.13801 -0.39298999 -0.23057 -0.22611 -0.14407 0.010511 -0.47389001 -0.15645 0.28601 -0.21772 -0.49535 0.022209 -0.23575 -0.22469001 -0.011578 0.52867001 -0.062309 ] ==>> Done !
4. Extract the feature of the whole sentense.
Given one sentense, such as "I Love Natural Language Processing", we can translate this sentense into a matrix representation. Specifically, we represent each word of the sentense with a vector which lies in a continuous space (as shown above). You can also see this blog: https://blog.csdn.net/fangqingan_java/article/details/50493948 to futher understand this process.
但是有如下的疑问：
1. 给定的句子，长短不一，每一个单词的维度固定，但是总的句子的长度不固定，怎么办？
只好用 padding 的方法了，那么，怎么进行 padding 呢？
For CNN architecture usually the input is (for each sentence) a vector of embedded words [GloVe(w0), GloVe(w1), ..., GloVe(wN)] of fixed length N.
So commonly you preset the maximum sentence length and just pad the tail of the input vector with zero vectors (just as you did).
Marking the beginning and the end of the sentence is more relevant for the RNN, where the sentence length is expected to be variable and processing is done sequentially.
Having said that, you can add a new dimension to the GloVe vector, setting it to zero for normal words, and arbitrarily to (say) 0.5 for BEGIN, and 1 for END and a random negative value for OOV.
As for unknown words, you should ;) encounter them very rear, otherwise you might consider training the embeddings by yourself.
那么，是什么意思呢？
对于 CNN 的网络来说，由于需要处理 网格状的 data，所以，需要将句子进行填充，以得到完整的 matrix，然后进行处理；
对于 RNN/LSTM 的网络来说，是可以按照时刻，顺序处理的，所以呢？就不需要 padding 了？
2. Here is another tutorial on Word2Vec using deep learning tools --- Keras.
from numpy import array from keras.preprocessing.text import one_hot from keras.preprocessing.sequence import pad_sequences from keras.models import Sequential from keras.layers import Dense from keras.layers import Flatten from keras.layers.embeddings import Embedding # define documents docs = ['Well done!', 'Good work', 'Great effort', 'nice work', 'Excellent!', 'Weak', 'Poor effort!', 'not good', 'poor work', 'Could have done better.'] # define class labels labels = array([1,1,1,1,1,0,0,0,0,0]) # integer encode the documents vocab_size = 50 encoded_docs = [one_hot(d, vocab_size) for d in docs] print("==>> encoded_docs: ") print(encoded_docs) # pad documents to a max length of 4 words max_length = 4 padded_docs = pad_sequences(encoded_docs, maxlen=max_length, padding='post') print("==>> padded_docs: ") print(padded_docs) # define the model model = Sequential() model.add(Embedding(vocab_size, 8, input_length=max_length)) model.add(Flatten()) model.add(Dense(1, activation='sigmoid')) # compile the model model.compile(optimizer='adam', loss='binary_crossentropy', metrics=['acc']) # summarize the model print("==>> model.summary()") print(model.summary()) # fit the model model.fit(padded_docs, labels, nb_epoch=100, verbose=0) # evaluate the model loss, accuracy = model.evaluate(padded_docs, labels, verbose=0) print("==>> the final Accuracy: ") print('Accuracy: %f' % (accuracy*100))
The output is:
Using TensorFlow backend. ==>> encoded_docs: [[26, 3], [18, 29], [39, 48], [34, 29], [23], [14], [3, 48], [40, 18], [3, 29], [40, 19, 3, 4]] ==>> padded_docs: [[26 3 0 0] [18 29 0 0] [39 48 0 0] [34 29 0 0] [23 0 0 0] [14 0 0 0] [ 3 48 0 0] [40 18 0 0] [ 3 29 0 0] [40 19 3 4]] WARNING:tensorflow:From /usr/local/lib/python2.7/dist-packages/keras/backend/tensorflow_backend.py:1047: calling reduce_prod (from tensorflow.python.ops.math_ops) with keep_dims is deprecated and will be removed in a future version. Instructions for updating: keep_dims is deprecated, use keepdims instead WARNING:tensorflow:From /usr/local/lib/python2.7/dist-packages/keras/backend/tensorflow_backend.py:1108: calling reduce_mean (from tensorflow.python.ops.math_ops) with keep_dims is deprecated and will be removed in a future version. Instructions for updating: keep_dims is deprecated, use keepdims instead ==>> model.summary() ____________________________________________________________________________________________________ Layer (type) Output Shape Param # Connected to ==================================================================================================== embedding_1 (Embedding) (None, 4, 8) 400 embedding_input_1[0][0] ____________________________________________________________________________________________________ flatten_1 (Flatten) (None, 32) 0 embedding_1[0][0] ____________________________________________________________________________________________________ dense_1 (Dense) (None, 1) 33 flatten_1[0][0] ==================================================================================================== Total params: 433 Trainable params: 433 Non-trainable params: 0 ____________________________________________________________________________________________________ None 2018-05-08 14:39:59.053198: I tensorflow/core/platform/cpu_feature_guard.cc:140] Your CPU supports instructions that this TensorFlow binary was not compiled to use: AVX2 FMA ==>> the final Accuracy: Accuracy: 89.999998
3. Maybe translate the sentence into vectors using skip-thought vectors is another goog choice. you can also check this tutorial from: http://www.cnblogs.com/wangxiaocvpr/p/7277025.html
4. But we still want to use GloVe to extract the vector of each word and concatenate them into a long vector.
1 import os
2 import numpy as np
3 from collections import OrderedDict
4 import re, pickle, collections, bcolz
5 import pdb
6
7 seq_home = '/dataset/language-dataset/'
8 seqlist_path = '/dataset/language-train-video-list.txt'
9 output_path = 'data/language-train-video-list.pkl'
10
11 def load_glove(loc):
12 return (bcolz.open(loc+'_glove_vecs.dat')[:],
13 pickle.load(open(loc+'_glove_words.pkl', 'rb')),
14 pickle.load(open(loc+'_glove_words_idx.pkl', 'rb')))
15
16 pre_trained_model_path = '/glove/glove.6B.200d.txt'
17
18 ###############################################
19 print('==>> Loading the glove.6B.200d.txt files.')
20 en_vecs, en_wv_word, en_wv_idx = load_glove(pre_trained_model_path)
21 en_w2v = {w: en_vecs[en_wv_idx[w]] for w in en_wv_word}
22 n_en_vec, dim_en_vec = en_vecs.shape
23
24
25 with open(seqlist_path,'r') as fp:
26 seq_list = fp.read().splitlines()
27
28 data = {}
29 for i,seq in enumerate(seq_list):
30 img_list = sorted([p for p in os.listdir(seq_home+seq+'/imgs/') if os.path.splitext(p)[1] == '.jpg']) ## image list
31 gt = np.loadtxt(seq_home+seq+'/groundtruth_rect.txt', delimiter=',') ## gt files
32 language_txt = open(seq_home+seq+'/language.txt', "rw+") ## natural language description files
33
34 line = language_txt.readline()
35 print("==>> language: %s" % (line))
36
37 gloveVector = []
38 test_txtName = seq_home+seq+'/glove_vector.txt'
39 f=file(test_txtName, "w")
40
41 word_list = line.split(' ')
42 for word_idx in range(len(word_list)):
43 current_word = word_list[word_idx]
44 current_GloVe_vector = en_w2v[current_word] ## vector dimension is: 200-D
45
46 gloveVector = np.concatenate((gloveVector, current_GloVe_vector), axis=0)
47 f.write(str(current_GloVe_vector))
48 f.write('\n')
49 f.close()
50
51 print(i)
52
53 assert len(img_list) == len(gt), "Lengths do not match!!"
54
55 if gt.shape[1]==8:
56 x_min = np.min(gt[:,[0,2,4,6]],axis=1)[:,None]
57 y_min = np.min(gt[:,[1,3,5,7]],axis=1)[:,None]
58 x_max = np.max(gt[:,[0,2,4,6]],axis=1)[:,None]
59 y_max = np.max(gt[:,[1,3,5,7]],axis=1)[:,None]
60 gt = np.concatenate((x_min, y_min, x_max-x_min, y_max-y_min),axis=1)
61
62 data[seq] = {'images':img_list, 'gt':gt, 'gloveVector': gloveVector}
63
64 with open(output_path, 'wb') as fp:
65 pickle.dump(data, fp, -1)
The Glove vector can be saved into a txt file for each video, as shown in following screenshots.
[ 0.19495 0.60610002 -0.077356 0.017301 -0.51370001 0.22421999 -0.80773002 0.022378 0.30256 1.06669998 -0.10918 0.57902998 0.23986 0.1691 0.0072384 0.42197999 -0.20458999 0.60271001 0.19188 -0.19616 0.33070001 3.20020008 -0.18104 0.20784 0.49511001 -0.42258999 0.022125 0.24379 0.16714001 -0.20909999 -0.12489 -0.51766998 -0.13569 -0.25979999 -0.17961 -0.47663 -0.89539999 -0.27138999 0.17746 0.45827001 0.21363001 0.22342999 -0.049342 0.34286001 -0.32315001 0.27469999 0.95993 -0.25979 0.21250001 -0.21373001 0.19809 0.15455 -0.48581001 0.38925001 0.33747 -0.27897999 0.19371 -0.45872 -0.054217 -0.24022999 0.59153003 0.12453 -0.21302 0.058223 -0.046671 -0.011614 -0.32025999 0.64120001 -0.28718999 0.035138 0.39287001 -0.030683 0.083529 -0.010964 -0.62427002 -0.13575 -0.38468999 0.11454 -0.61037999 0.12594999 -0.17633 -0.21415 -0.37013999 0.21763 0.055366 -0.25242001 -0.45475999 -0.28105 0.18911 -1.58539999 0.64841002 -0.34621999 0.59254003 -0.39034 -0.44258001 0.20562001 0.44395 0.23019999 -0.35018 -0.36090001 0.62993002 -0.34698999 -0.31964999 -0.17361 0.51714998 0.68493003 -0.15587001 1.42550004 -0.94313997 0.031951 -0.26210001 -0.089544 0.22437 -0.050374 0.035414 -0.070493 0.17037 -0.38598001 0.0095626 0.26363 0.72125 -0.13797 0.70602 -0.50839001 -0.49722001 -0.48706001 0.16254 0.025619 0.33572 -0.64160001 -0.32541999 0.21896 0.05331 0.082011 0.12038 0.088849 -0.04651 -0.085435 0.036835 -0.14695001 -0.25841001 -0.043812 0.053209 -0.48954999 1.73950005 0.99014997 0.09395 -0.20236 -0.050135 0.18337999 0.22713999 0.83146 -0.30974001 0.51995999 0.068264 -0.28237 -0.30096999 -0.031014 0.024615 0.4294 -0.085036 0.051913 0.31251001 -0.34443 -0.085145 0.024975 0.0082017 0.17241 -0.66000998 0.0058962 -0.055387 -0.22315 0.35721999 -0.18962 0.25819999 -0.24685 -0.79571998 -0.09436 -0.10271 0.13713001 1.48660004 -0.57113999 -0.52649999 -0.25181001 -0.40792 -0.18612 0.040009 0.11557 0.017987 0.27149001 -0.14252 -0.087756 0.15196 0.064926 0.01651 -0.25334001 0.27437001 0.24246 0.018494 0.22473 ] [ 4.32489991e-01 5.34709990e-01 -1.83240008e-02 1.56369999e-01 6.69689998e-02 7.63140023e-02 -7.16499984e-01 2.41520002e-01 1.70919999e-01 8.96220028e-01 -2.00739995e-01 1.88370004e-01 4.53570008e-01 -1.86419994e-01 -3.41060013e-01 -3.10570002e-02 -4.90720011e-02 6.55830026e-01 -7.43409991e-03 1.54770002e-01 1.55849993e-01 3.11129999e+00 -4.12149996e-01 1.89439997e-01 1.86639994e-01 -2.00540006e-01 6.63070008e-02 3.90180014e-02 6.06740005e-02 -3.17880005e-01 1.82290003e-02 -6.80719972e-01 -1.33360000e-02 2.32099995e-01 -8.23459998e-02 -2.68420011e-01 -3.51870000e-01 -2.03860000e-01 6.69749975e-02 3.02689999e-01 1.84330001e-01 1.44429997e-01 1.38980001e-01 1.93629991e-02 -4.09139991e-01 2.24950001e-01 8.02020013e-01 -3.05290014e-01 2.22399995e-01 -6.10979982e-02 3.48910004e-01 1.13349997e-01 -4.14329991e-02 7.31640011e-02 3.17719996e-01 -2.59590000e-01 3.17759991e-01 -1.78080007e-01 2.12380007e-01 1.07529998e-01 4.37519997e-01 -1.81329995e-01 -1.53909996e-01 7.27130026e-02 4.14730012e-01 2.73079991e-01 -1.06940001e-01 7.45599985e-01 3.02549988e-01 1.49999997e-02 6.33790016e-01 1.03399999e-01 2.18640000e-01 -7.43470015e-03 -7.09949970e-01 4.60739993e-02 -8.34990025e-01 1.84519999e-02 -2.75200009e-01 -2.48000007e-02 -1.21320002e-01 3.29720005e-02 1.07320003e-01 2.18070000e-01 2.79079992e-02 -2.99050003e-01 -3.06030005e-01 -3.17970008e-01 2.10580006e-01 -1.16180003e+00 4.57720011e-01 -2.60210007e-01 3.76190007e-01 -2.16749996e-01 -1.68540001e-01 1.77729994e-01 -1.46709993e-01 2.67820001e-01 -5.86109996e-01 -4.98769999e-01 3.83679986e-01 2.51020014e-01 -3.90020013e-01 -6.41409993e-01 5.34929991e-01 2.76479989e-01 -4.32830006e-01 1.61350000e+00 -1.58399999e-01 -2.03970000e-01 -2.18089998e-01 -9.74659983e-04 -7.48440027e-02 3.78959998e-02 -4.03719991e-01 -1.74640000e-01 4.11469996e-01 -2.80400008e-01 -3.42570007e-01 5.72329983e-02 7.18890011e-01 -1.30419999e-01 5.75810015e-01 -3.27690005e-01 7.75609985e-02 -2.66130000e-01 4.82820012e-02 1.24679998e-01 1.29319996e-01 7.65379965e-02 -3.85809988e-01 3.72209996e-01 -1.67280003e-01 1.86330006e-01 -2.20280007e-01 3.27170014e-01 1.74089998e-01 -2.51549989e-01 -5.47109991e-02 -2.48549998e-01 -2.24260002e-01 -3.80129993e-01 5.53080022e-01 -1.52050003e-01 1.12820005e+00 3.70869994e-01 -4.62579988e-02 -3.56009990e-01 -1.72120005e-01 1.23460002e-01 6.58179998e-01 5.05930007e-01 -4.99610007e-01 1.97339997e-01 1.37759998e-01 3.98500008e-04 -1.38300002e-01 -6.67430013e-02 -7.20809996e-02 4.00720000e-01 -3.77620012e-01 -9.86199975e-02 2.04119995e-01 -3.42869997e-01 -2.01509997e-01 -1.18270002e-01 -1.60109997e-01 1.63570002e-01 -3.79029989e-01 -2.45529994e-01 -5.31699993e-02 -4.19999994e-02 1.82099998e-01 2.23959997e-01 4.50800002e-01 -2.03030005e-01 -6.16330028e-01 -2.02739999e-01 2.18419999e-01 3.66939992e-01 7.68440008e-01 -2.92189986e-01 -9.80089977e-02 -2.92939991e-01 -1.93189994e-01 1.45720005e-01 2.45150000e-01 -7.11840019e-02 3.97929996e-01 -3.33019998e-03 -4.01179999e-01 1.39760002e-01 8.27540010e-02 -1.29600003e-01 -3.05500001e-01 8.98869988e-03 2.26559997e-01 3.21410000e-01 -4.29780006e-01 4.74779993e-01] [ 5.73459983e-01 5.41700006e-01 -2.34770000e-01 -3.62399995e-01 4.03699994e-01 1.13860004e-01 -4.49330002e-01 -3.09909999e-01 -5.34110004e-03 5.84259987e-01 -2.59559993e-02 4.93930012e-01 -3.72090004e-02 -2.84280002e-01 9.76959988e-02 -4.89069998e-01 2.60269996e-02 3.76489997e-01 5.77879995e-02 -4.68070000e-01 8.12880024e-02 3.28250003e+00 -6.36900008e-01 3.79559994e-01 3.81670007e-03 9.36070010e-02 -1.28549993e-01 1.73800007e-01 1.05219997e-01 2.86480010e-01 2.10889995e-01 -4.70759988e-01 2.77330000e-02 -1.98029995e-01 7.63280019e-02 -8.46289992e-01 -7.97079980e-01 -3.87430012e-01 -3.04220002e-02 -2.68489987e-01 4.85850006e-01 1.28950000e-01 3.83540004e-01 3.87219995e-01 -3.85239989e-01 1.90750003e-01 4.89980012e-01 1.32780001e-01 1.07920002e-02 2.67699987e-01 1.78120002e-01 -1.14330001e-01 -3.34939986e-01 8.73059988e-01 7.58750021e-01 -3.03779989e-01 -1.56259999e-01 1.20850001e-03 2.33219996e-01 2.79529989e-01 -1.84939995e-01 -1.41460001e-01 -1.89689994e-01 -3.83859985e-02 3.58740002e-01 6.55129999e-02 6.05649985e-02 6.63389981e-01 -8.32519978e-02 6.51630014e-02 5.17610013e-01 1.61709994e-01 4.60110009e-01 1.63880005e-01 -1.23989999e-01 3.11219990e-01 -1.54119998e-01 -1.09169997e-01 -4.25509989e-01 1.14179999e-01 2.51370013e-01 -5.61579987e-02 -2.59270012e-01 2.81630009e-01 -1.80939995e-02 1.60650000e-01 -4.85060006e-01 -9.89030004e-01 2.50220001e-01 -1.67359993e-01 4.14739996e-01 1.77010000e-01 4.24070001e-01 1.10880002e-01 -1.83599994e-01 -1.24100000e-01 -3.47799987e-01 9.90779996e-02 -2.23810002e-01 -1.12450004e-01 -2.11559996e-01 3.07060010e-03 -2.36070007e-01 2.72610001e-02 3.64300013e-01 3.99219990e-02 -1.83689997e-01 1.22660005e+00 -7.76400030e-01 -6.62249982e-01 1.57239996e-02 -1.49690002e-01 8.46489966e-02 2.68139988e-01 -1.67649999e-01 -3.19420010e-01 2.84940004e-01 -7.00000003e-02 1.20099997e-02 -1.22189999e-01 5.63099980e-01 -3.19999993e-01 5.01089990e-01 -1.02090001e-01 4.65750009e-01 -7.15420008e-01 1.72930002e-01 5.82589984e-01 7.83839971e-02 -3.38440016e-02 -2.51289994e-01 3.65029991e-01 3.15780006e-02 -6.57779992e-01 5.47499992e-02 8.71890008e-01 1.24550000e-01 -4.58770007e-01 -2.69650012e-01 -4.67790008e-01 -2.85780011e-03 1.78100005e-01 6.39689982e-01 1.39950007e-01 9.75960016e-01 1.18359998e-01 -6.39039993e-01 -1.54159993e-01 6.52619973e-02 2.43290007e-01 6.64759994e-01 2.50690013e-01 -1.02519996e-01 -3.28390002e-01 -8.55590031e-02 -1.27739999e-02 -1.94309995e-01 5.61389983e-01 -3.57329994e-01 -2.03439996e-01 -1.24130003e-01 -3.44309986e-01 -2.32960001e-01 -2.11870000e-01 8.53869990e-02 7.00630024e-02 -1.98029995e-01 -2.60230005e-02 -3.90370011e-01 8.00019979e-01 4.05770004e-01 -7.98629969e-02 3.52629989e-01 -3.40429991e-01 3.96759987e-01 2.28619993e-01 -3.50279987e-01 -4.73439991e-01 5.97419977e-01 -1.16570003e-01 1.05519998e+00 -4.15699989e-01 -8.05519968e-02 -5.65709993e-02 -1.66219994e-01 1.92739993e-01 -9.51749980e-02 -2.07810000e-01 1.56200007e-01 5.02309985e-02 -2.79150009e-01 4.37420011e-01 -3.12370002e-01 1.31940007e-01 -3.32780004e-01 1.88769996e-01 -2.34219998e-01 5.44179976e-01 -2.30690002e-01 3.49469990e-01] [ -1.43299997e-01 -6.54269993e-01 -4.71419990e-01 -2.22780004e-01 -1.74260005e-01 6.43490016e-01 -7.60049999e-01 -8.42769966e-02 2.19850004e-01 -4.24650013e-01 -2.56919991e-02 -3.27909999e-02 -3.98149997e-01 -1.26000000e-02 6.31869972e-01 -6.31640017e-01 -2.70480007e-01 2.09920004e-01 5.72669983e-01 8.88589993e-02 2.24289998e-01 1.78419995e+00 8.73669982e-01 -2.23949999e-01 4.91869986e-01 6.86379969e-02 6.80689991e-01 1.85680002e-01 -2.85780013e-01 -1.06030002e-01 -3.07460010e-01 1.06180003e-02 -1.62290007e-01 -3.62700000e-02 8.02920014e-02 -2.17669994e-01 -7.92110026e-01 -7.06340015e-01 -2.89400011e-01 -1.51869997e-01 3.09009999e-01 -1.41849995e-01 -4.18920010e-01 -1.66299999e-01 -1.68510005e-01 1.04549997e-01 -3.49110007e-01 -1.21399999e+00 3.21929991e-01 1.88219994e-01 -3.50019991e-01 -2.23629996e-01 3.85280013e-01 1.40709996e-01 1.34890005e-01 3.89420003e-01 1.82109997e-01 1.32369995e-01 -2.60610014e-01 5.10110021e-01 1.26430005e-01 -2.94450015e-01 -1.34350002e-01 -1.44600004e-01 1.98740005e-01 -3.41720015e-01 2.40710005e-01 -6.08640015e-02 8.60409975e-01 8.92559998e-04 3.46020013e-01 -2.53780007e-01 2.53749996e-01 1.20920002e-01 -3.82739991e-01 1.55090000e-02 -6.41290024e-02 2.54729986e-01 -1.78489998e-01 2.37949997e-01 3.26880008e-01 2.03940004e-01 -5.65020025e-01 3.87109995e-01 -7.00220019e-02 -4.63999987e-01 -1.98210001e-01 -7.27079988e-01 5.36920011e-01 -9.09250021e-01 -2.36699998e-01 -6.20739982e-02 2.38769993e-01 3.24860007e-01 5.51190019e-01 -4.07079995e-01 1.40870005e-01 2.83569992e-01 -1.33379996e-01 -3.99740010e-01 -1.66620001e-01 4.43010002e-01 -6.85970008e-01 1.54200003e-01 6.58490002e-01 2.95630004e-02 2.61550009e-01 1.10969996e+00 -6.78719997e-01 -1.91960007e-01 7.45159984e-02 -3.53740007e-01 -7.15600014e-01 4.14079994e-01 -3.37280005e-01 2.05780007e-02 2.17340007e-01 6.48889989e-02 -3.51080000e-01 -2.44440004e-01 -1.98780000e-01 3.02210003e-01 -1.32180005e-01 -4.86490011e-01 -3.55599999e-01 2.83119995e-02 7.04949975e-01 7.35610008e-01 5.98779976e-01 -3.94160002e-01 -3.96349996e-01 2.55739987e-01 1.30939996e+00 -6.69730008e-02 4.72090006e-01 3.20030004e-01 -2.39020005e-01 1.87910005e-01 2.14310005e-01 3.10609996e-01 -7.22400010e-01 -2.59559989e-01 7.50709996e-02 5.45759976e-01 1.33249998e+00 5.97169995e-01 -7.56640017e-01 2.30540007e-01 3.45209986e-01 8.61710012e-01 5.65129995e-01 1.33000001e-01 4.27370012e-01 -4.36790008e-03 -5.15600026e-01 3.94329995e-01 -1.58559993e-01 -1.29820004e-01 -5.47240004e-02 -1.01789999e+00 6.39170036e-02 -6.47679985e-01 -7.53360018e-02 1.49210006e-01 5.53300023e-01 -6.24869987e-02 -2.01309994e-01 1.70489997e-01 -3.22829992e-01 2.43039995e-01 2.64449995e-02 -9.52059999e-02 3.80879998e-01 -3.69230002e-01 4.54270005e-01 -3.31169993e-01 -1.93750001e-02 1.59300007e-02 1.85659993e-02 -5.27350008e-01 1.05949998e+00 -4.31519985e-01 2.57169992e-01 -9.44539979e-02 7.53460005e-02 -2.06599995e-01 3.21940005e-01 -1.23549998e-02 4.72169995e-01 2.44020000e-01 -3.63059998e-01 1.71650007e-01 -3.50499988e-01 1.40990004e-01 1.18069999e-01 3.65170002e-01 -1.65889993e-01 -3.93799990e-02 -3.70829999e-01 2.03909993e-01] [ 2.41689995e-01 -3.45340014e-01 -2.23069996e-01 -1.29069996e+00 2.52849996e-01 -5.51280022e-01 -8.03359970e-02 -8.17670021e-03 3.11360002e-01 -4.51009989e-01 2.46610001e-01 3.64410013e-01 9.43359971e-01 -3.54200006e-02 7.80480027e-01 -3.97650003e-01 3.11250001e-01 -1.77430004e-01 -4.19889987e-01 -3.78149986e-01 6.72299981e-01 3.17160010e+00 3.24960016e-02 -3.16400006e-02 5.80680013e-01 -4.44579989e-01 -5.56120016e-02 1.80519998e-01 2.85719991e-01 9.58700031e-02 2.14369997e-01 4.97310013e-02 1.87199995e-01 1.19139999e-01 2.74080001e-02 -8.06079984e-01 -3.08349997e-01 -8.97369981e-01 -1.97720006e-01 2.67409999e-02 -3.87650013e-01 1.16590001e-01 -2.01100007e-01 2.01010004e-01 -7.91329965e-02 -5.09539992e-02 6.01889985e-03 3.34699988e-01 -2.11180001e-01 7.40419999e-02 -2.81410009e-01 -5.96150011e-02 -3.52959991e-01 6.47480011e-01 5.39080016e-02 -3.13760012e-01 -3.66219997e-01 -2.77550012e-01 2.26760004e-02 4.88109998e-02 1.43120006e-01 -1.85800001e-01 -5.69639981e-01 -5.41190028e-01 1.86159998e-01 1.88539997e-01 2.75209993e-01 -1.78350002e-01 -3.74379992e-01 1.21090002e-01 1.86099997e-03 -9.21269972e-03 1.01860002e-01 9.80810001e-02 -3.72449994e-01 6.64150000e-01 5.73659986e-02 -4.38450009e-01 -4.05250013e-01 -5.59589982e-01 -1.13430001e-01 -5.49870014e-01 -2.63209999e-01 -2.84709990e-01 1.44109994e-01 1.03600003e-01 -3.21980000e-01 -2.15299994e-01 9.86679971e-01 -4.19369996e-01 3.11899990e-01 3.33799988e-01 1.60180002e-01 3.31369996e-01 -2.54939999e-02 -3.78799997e-02 -1.20480001e-01 -1.21749997e-01 9.47659984e-02 2.61579990e-01 2.99309995e-02 -2.96249986e-01 4.34009999e-01 -3.65360007e-02 -4.28519994e-01 -3.96380007e-01 -2.49730006e-01 1.10179996e+00 -2.28119999e-01 2.43239999e-01 8.38569999e-02 -4.88810003e-01 -2.13159993e-01 4.02490012e-02 -4.05160010e-01 -1.24109998e-01 -1.97280005e-01 -7.36959994e-01 -4.45820004e-01 -4.49919999e-01 -7.37069994e-02 -1.64539993e-01 1.60109997e-01 -4.19519991e-01 4.14169997e-01 -5.72770000e-01 5.13670027e-01 8.11180025e-02 2.92879995e-02 3.53089988e-01 -1.08029999e-01 1.37020007e-01 3.88280004e-01 -2.88129985e-01 6.82030022e-01 1.74119994e-01 -8.85400027e-02 -2.97850013e-01 -2.84550011e-01 1.35700002e-01 1.52319998e-01 2.20369995e-01 6.88120008e-01 -1.30999997e-01 1.81970000e+00 -4.51530010e-01 3.95529985e-01 -6.08169973e-01 3.36979985e-01 1.10900000e-01 -2.84759998e-01 3.04610014e-01 -2.16059998e-01 -5.81110008e-02 4.41720009e-01 -2.42310002e-01 -1.27580002e-01 -4.87929992e-02 -1.54949993e-01 -8.54910016e-01 9.29120034e-02 -6.08090013e-02 2.90730000e-02 -3.87349993e-01 -7.08530024e-02 -6.59749985e-01 -3.81570011e-01 5.01699984e-01 -7.35599995e-01 4.15210009e-01 2.13280007e-01 -3.37790012e-01 6.69019997e-01 4.24860001e-01 -1.21480003e-01 -1.06260004e-02 1.27450004e-01 -1.35609999e-01 2.34229997e-01 3.51099998e-01 1.28410006e+00 1.29820004e-01 2.13569999e-01 3.28570008e-01 1.65669993e-01 -2.14579999e-01 -4.42750007e-01 3.28500003e-01 1.80010006e-01 6.48650005e-02 -3.58799994e-01 -1.42259998e-02 3.11250001e-01 -2.20489994e-01 3.28290015e-02 3.85250002e-01 -1.05120003e-01 2.78010011e-01 -1.01709999e-01 -7.15209991e-02] [ 5.24869978e-01 -1.19410001e-01 -2.02419996e-01 -6.23929977e-01 -1.53799996e-01 -6.63369969e-02 -3.68499994e-01 2.86490005e-02 1.37950003e-01 -5.87819993e-01 6.02090001e-01 2.60540005e-02 7.07889974e-01 1.20329998e-01 -1.74310002e-02 4.03360009e-01 -3.19680005e-01 -2.50609994e-01 1.60889998e-01 2.47649997e-01 7.79359996e-01 2.74070001e+00 1.19589999e-01 -2.67529994e-01 -3.82809997e-01 -3.36580008e-01 1.41039997e-01 -4.65480000e-01 -8.92079994e-02 2.22540006e-01 -3.60740013e-02 -7.10140020e-02 6.23199999e-01 3.22770000e-01 4.15650010e-01 -3.68530005e-02 -5.82859993e-01 -6.26510024e-01 -3.26169990e-02 2.74789989e-01 -2.66950011e-01 5.27690016e-02 -1.09500003e+00 -1.99760008e-03 -7.49390006e-01 -1.89999994e-02 -1.87619999e-01 -5.19330025e-01 1.71590000e-01 4.40690011e-01 1.90789998e-01 -4.57339995e-02 -2.43069995e-02 2.32710004e-01 1.02179997e-01 -6.02790006e-02 -2.63680011e-01 -1.49090007e-01 4.33890015e-01 -8.51420015e-02 -6.61419988e-01 -6.23379983e-02 -1.37920007e-01 4.54079986e-01 -1.51400000e-01 1.14929996e-01 5.48650026e-01 2.82370001e-01 -2.55129993e-01 1.11149997e-01 -8.47479999e-02 -9.66319963e-02 4.88200009e-01 7.48070002e-01 -6.74910009e-01 3.21949989e-01 -8.98810029e-01 -3.58159989e-01 8.54640007e-02 -1.67390004e-01 -1.60109997e-01 -1.06349997e-01 -2.96649992e-01 -3.03389996e-01 2.29310006e-01 6.51580021e-02 1.70359999e-01 -2.08039999e-01 -1.12930000e-01 -1.58399999e-01 -1.88710004e-01 5.38689971e-01 -3.58050019e-02 3.26770008e-01 -3.18859994e-01 2.01079994e-02 -1.15209997e-01 -3.67400013e-02 5.97510003e-02 9.09829974e-01 1.55420005e-01 1.63939998e-01 3.57499987e-01 -4.75789994e-01 -3.53519991e-02 -1.30729997e+00 2.40319997e-01 1.24170005e+00 4.63180006e-01 -2.90019996e-02 6.63789988e-01 -2.20129997e-01 6.57410026e-01 -2.34209999e-01 -7.43539989e-01 8.95029977e-02 -1.41269997e-01 3.54860008e-01 -6.87049985e-01 -9.53029990e-02 -1.59720004e-01 -4.36020009e-02 5.39509989e-02 -1.33180007e-01 4.47530001e-01 1.41289994e-01 2.71119997e-02 1.18179999e-01 2.02419996e-01 -5.98179996e-01 -6.83469996e-02 -1.58150002e-01 2.42929995e-01 8.88189971e-02 -1.16219997e-01 -3.82679999e-01 1.30649999e-01 -3.00619990e-01 -8.46930027e-01 1.90159995e-02 6.41009986e-01 -8.42439979e-02 7.83680022e-01 -2.63660014e-01 1.98839998e+00 -7.38060027e-02 7.21970022e-01 4.84739989e-02 1.62070006e-01 3.40840012e-01 -1.85289994e-01 2.85659999e-01 -2.35440001e-01 -2.69939989e-01 -5.27239978e-01 2.74859995e-01 6.56930029e-01 2.20569998e-01 -6.80390000e-01 3.59100014e-01 2.51430005e-01 7.24399984e-02 2.58690000e-01 3.22510004e-01 3.57239991e-02 -3.88200015e-01 1.18890002e-01 -2.90010005e-01 -6.32229984e-01 -1.48519993e-01 -1.21040002e-01 -1.74060002e-01 2.48679996e-01 4.74539995e-01 -2.50710011e-01 -5.23679972e-01 -1.82730004e-01 -8.30529988e-01 6.15109980e-01 -3.84620011e-01 9.74919975e-01 4.49510008e-01 -5.42559981e-01 -3.18340003e-01 -5.07809997e-01 -2.73739994e-02 -1.26379997e-01 5.42479992e-01 4.85489994e-01 8.55289996e-02 -1.42509997e-01 8.77849981e-02 2.47189999e-01 -2.38490000e-01 6.14570007e-02 -4.23129983e-02 3.57470006e-01 9.73920003e-02 -3.58500004e-01 1.70269996e-01] [ 7.54669979e-02 -2.92360008e-01 -2.60369986e-01 -2.81670004e-01 1.60970002e-01 -1.94720000e-01 -2.82059997e-01 4.91409987e-01 -4.11189999e-03 9.58790034e-02 3.33959997e-01 1.68740004e-02 2.06159994e-01 -1.57429993e-01 -1.61379993e-01 2.15379998e-01 -1.48359999e-01 2.54599992e-02 -4.36809987e-01 -6.36610016e-02 5.52609980e-01 3.08189988e+00 -1.17839999e-01 -4.61309999e-01 -2.76089996e-01 2.09480003e-01 -2.05440000e-01 -5.71550012e-01 3.34479988e-01 1.59130007e-01 2.54360004e-03 1.80040002e-01 1.34719998e-01 -9.74040031e-02 3.55369985e-01 -4.74280000e-01 -7.92569995e-01 -5.44179976e-01 2.42999997e-02 6.35990024e-01 1.23369999e-01 -1.29130006e-01 -2.65650004e-01 -2.49569997e-01 -5.21990001e-01 -4.05229986e-01 4.84030008e-01 1.83729995e-02 2.30389997e-01 6.21379986e-02 -1.92919999e-01 2.95060009e-01 -3.57930005e-01 1.67019993e-01 3.18679988e-01 -3.60540003e-01 -1.09779999e-01 -1.56320006e-01 4.59210008e-01 9.64749977e-02 -3.77000004e-01 -7.76150003e-02 -4.88990009e-01 2.05750003e-01 5.05429983e-01 5.34190014e-02 -2.59779990e-01 5.10420024e-01 9.75209996e-02 3.26059997e-01 1.43539995e-01 2.27149995e-03 4.86149997e-01 4.69379991e-01 -4.11240011e-01 -1.71480000e-01 -3.97439986e-01 -2.89009988e-01 -1.77560002e-01 3.70009989e-02 3.48300010e-01 1.59339994e-01 -7.42810011e-01 1.88970000e-01 4.36850004e-02 5.72080016e-01 -6.70159996e-01 -4.39470001e-02 -2.83360004e-01 -3.19959998e-01 -2.04040006e-01 -8.78980011e-02 -1.57240003e-01 2.18180008e-02 -5.67569971e-01 6.32960021e-01 -1.00970000e-01 -6.55760020e-02 5.82689978e-03 3.30349989e-02 3.97830009e-01 -3.11659992e-01 -6.10889971e-01 2.75590003e-01 1.00079998e-01 -4.19900000e-01 6.35599997e-03 1.87170005e+00 3.14729989e-01 -3.60040009e-01 8.13839972e-01 -2.17099994e-01 -1.84589997e-02 -2.26319999e-01 1.45850003e-01 -1.43500000e-01 -4.14239988e-02 5.59740007e-01 -6.67519987e-01 -2.19589993e-01 1.90109998e-01 3.30150008e-01 6.12900019e-01 4.67709988e-01 4.20260012e-01 -5.28190017e-01 2.31650006e-02 3.29100005e-02 4.73060012e-01 1.40060000e-02 -1.73960000e-01 -4.43619996e-01 4.13769990e-01 -2.06790000e-01 3.92830014e-01 3.02109987e-01 7.31339976e-02 4.21640016e-02 -9.27100003e-01 -4.76139992e-01 2.43100002e-01 -1.33790001e-01 -2.22379997e-01 -4.14570011e-02 1.58500004e+00 3.74810010e-01 2.59939991e-02 -2.42719993e-01 3.05779994e-01 1.46870002e-01 1.16659999e-01 -2.94179991e-02 -7.83390030e-02 -2.25119993e-01 1.33149996e-01 -6.48420006e-02 -2.86870003e-01 -1.05600003e-02 -3.46679986e-01 4.21449989e-02 -6.00409985e-01 8.24810028e-01 3.10220003e-01 1.64890006e-01 -7.29210004e-02 1.93939999e-01 -9.84980017e-02 -2.03830004e-02 -4.09090012e-01 -1.04039997e-01 1.91689998e-01 -1.59689993e-01 3.80259991e-01 6.28019989e-01 2.59499997e-01 -3.33669990e-01 -7.33330011e-01 -4.07429993e-01 6.84230030e-01 -6.63380027e-02 5.04360020e-01 -2.89829999e-01 -3.90859991e-01 -4.59310003e-02 -2.66240001e-01 -1.67699993e-01 -1.50370002e-01 1.48279995e-01 -2.81430006e-01 -1.70870006e-01 -2.55760014e-01 -5.62830009e-02 -1.66500002e-01 3.51060003e-01 4.10320014e-02 2.73110002e-01 3.00200004e-02 1.64649993e-01 -8.41889977e-02 5.75059988e-02] [ 5.46909988e-01 -7.55890012e-01 -9.20799971e-01 -8.20680022e-01 1.48800001e-01 -1.32200003e-01 2.49499991e-03 5.39979994e-01 -3.12929988e-01 -2.12009996e-02 -2.96559989e-01 1.51110003e-02 -3.76980007e-02 -4.07290012e-01 3.36799994e-02 -2.90179998e-01 -5.64790010e-01 6.47960007e-01 3.96770000e-01 -1.91139996e-01 6.98249996e-01 1.61090004e+00 8.42899978e-01 -1.45980000e-01 -1.13940001e-01 -5.39680004e-01 -7.22540021e-01 -1.47310004e-01 -6.78520024e-01 1.93159997e-01 4.17819992e-03 -3.96059990e-01 8.99320021e-02 3.30909997e-01 -4.09649983e-02 4.59470004e-01 -3.07020009e-01 -2.61469990e-01 5.39489985e-01 6.60130024e-01 -5.68049997e-02 -3.70750010e-01 -1.46880001e-03 -1.48770005e-01 2.13000000e-01 -1.36710003e-01 5.12350023e-01 -5.27249992e-01 2.73449998e-02 3.00159991e-01 4.59829986e-01 3.65709990e-01 2.23989993e-01 1.47589996e-01 -9.97050032e-02 2.26520002e-01 -5.95120013e-01 -4.34839994e-01 -5.26629984e-02 -3.05240005e-01 -1.91280007e-01 -4.08630013e-01 -2.83719987e-01 6.89159989e-01 -7.18529999e-01 2.89909989e-01 -1.31809995e-01 -8.75229985e-02 2.70280004e-01 3.31449993e-02 1.93159997e-01 -2.50629991e-01 1.13150001e-01 4.88560013e-02 -4.41439986e-01 -8.88149977e-01 -3.45840007e-01 -5.47370017e-01 2.13919997e-01 1.46469995e-01 -3.76450002e-01 5.63120008e-01 4.44130003e-01 -6.05080016e-02 5.85870028e-01 4.50850010e-01 -1.27309993e-01 6.86409995e-02 3.78729999e-01 -7.15900004e-01 2.72450000e-01 2.16839999e-01 -5.09050012e-01 2.82240003e-01 5.28559983e-01 2.97140002e-01 -3.33429992e-01 -1.62200004e-01 -3.97210002e-01 3.37130010e-01 6.61469996e-01 -4.09590006e-01 -1.97520003e-01 -8.77860010e-01 -4.36379999e-01 1.79399997e-01 3.39769991e-03 9.73770022e-01 7.32789993e-01 -3.44499983e-02 -4.76709992e-01 -4.99819994e-01 4.97110009e-01 -7.77419984e-01 -3.60000014e-01 -1.58899993e-01 1.19869998e-02 -1.39750004e-01 -5.91250002e-01 -1.58280000e-01 9.19710025e-02 4.62440014e-01 -1.17980000e-02 8.29930007e-01 7.96280026e-01 -5.64369977e-01 -1.02080004e-02 -3.86090010e-01 -6.74889982e-01 -3.42109986e-02 1.99829996e-01 2.01010004e-01 7.43009984e-01 -3.58200014e-01 1.07579999e-01 -7.77289987e-01 3.24770004e-01 -7.43889987e-01 -7.90560007e-01 4.95799989e-01 3.70090008e-01 -1.55349998e-02 1.20260000e+00 -6.90900028e-01 1.49619997e+00 4.31109995e-01 2.64299989e-01 -5.99219978e-01 -4.51359987e-01 7.66879976e-01 -1.43020004e-02 1.43340006e-01 1.19379997e-01 -5.56930006e-01 -9.95339990e-01 8.20089996e-01 -3.02410007e-01 9.95709971e-02 -4.92139995e-01 2.25950003e-01 5.18890023e-01 6.51669979e-01 -7.13440031e-02 3.06109991e-03 -2.95789987e-01 -4.40490007e-01 -1.61009997e-01 2.52220005e-01 -3.50809991e-01 -7.96020031e-02 4.33560014e-01 -1.43889993e-01 -1.26969993e-01 1.01450002e+00 -6.24639988e-02 -5.60909986e-01 -6.52670026e-01 1.86299995e-01 -3.76190007e-01 -2.40750000e-01 2.50979990e-01 -2.06090003e-01 -8.57209980e-01 -5.84480017e-02 -1.23539999e-01 -7.35830009e-01 -4.83509988e-01 2.75110006e-01 -1.10220000e-01 3.17330003e-01 1.87999994e-01 1.05719995e+00 5.85460007e-01 3.13740000e-02 2.11569995e-01 1.13799997e-01 4.65829998e-01 -3.05020005e-01 -4.51409996e-01 2.87349999e-01] [ 2.41689995e-01 -3.45340014e-01 -2.23069996e-01 -1.29069996e+00 2.52849996e-01 -5.51280022e-01 -8.03359970e-02 -8.17670021e-03 3.11360002e-01 -4.51009989e-01 2.46610001e-01 3.64410013e-01 9.43359971e-01 -3.54200006e-02 7.80480027e-01 -3.97650003e-01 3.11250001e-01 -1.77430004e-01 -4.19889987e-01 -3.78149986e-01 6.72299981e-01 3.17160010e+00 3.24960016e-02 -3.16400006e-02 5.80680013e-01 -4.44579989e-01 -5.56120016e-02 1.80519998e-01 2.85719991e-01 9.58700031e-02 2.14369997e-01 4.97310013e-02 1.87199995e-01 1.19139999e-01 2.74080001e-02 -8.06079984e-01 -3.08349997e-01 -8.97369981e-01 -1.97720006e-01 2.67409999e-02 -3.87650013e-01 1.16590001e-01 -2.01100007e-01 2.01010004e-01 -7.91329965e-02 -5.09539992e-02 6.01889985e-03 3.34699988e-01 -2.11180001e-01 7.40419999e-02 -2.81410009e-01 -5.96150011e-02 -3.52959991e-01 6.47480011e-01 5.39080016e-02 -3.13760012e-01 -3.66219997e-01 -2.77550012e-01 2.26760004e-02 4.88109998e-02 1.43120006e-01 -1.85800001e-01 -5.69639981e-01 -5.41190028e-01 1.86159998e-01 1.88539997e-01 2.75209993e-01 -1.78350002e-01 -3.74379992e-01 1.21090002e-01 1.86099997e-03 -9.21269972e-03 1.01860002e-01 9.80810001e-02 -3.72449994e-01 6.64150000e-01 5.73659986e-02 -4.38450009e-01 -4.05250013e-01 -5.59589982e-01 -1.13430001e-01 -5.49870014e-01 -2.63209999e-01 -2.84709990e-01 1.44109994e-01 1.03600003e-01 -3.21980000e-01 -2.15299994e-01 9.86679971e-01 -4.19369996e-01 3.11899990e-01 3.33799988e-01 1.60180002e-01 3.31369996e-01 -2.54939999e-02 -3.78799997e-02 -1.20480001e-01 -1.21749997e-01 9.47659984e-02 2.61579990e-01 2.99309995e-02 -2.96249986e-01 4.34009999e-01 -3.65360007e-02 -4.28519994e-01 -3.96380007e-01 -2.49730006e-01 1.10179996e+00 -2.28119999e-01 2.43239999e-01 8.38569999e-02 -4.88810003e-01 -2.13159993e-01 4.02490012e-02 -4.05160010e-01 -1.24109998e-01 -1.97280005e-01 -7.36959994e-01 -4.45820004e-01 -4.49919999e-01 -7.37069994e-02 -1.64539993e-01 1.60109997e-01 -4.19519991e-01 4.14169997e-01 -5.72770000e-01 5.13670027e-01 8.11180025e-02 2.92879995e-02 3.53089988e-01 -1.08029999e-01 1.37020007e-01 3.88280004e-01 -2.88129985e-01 6.82030022e-01 1.74119994e-01 -8.85400027e-02 -2.97850013e-01 -2.84550011e-01 1.35700002e-01 1.52319998e-01 2.20369995e-01 6.88120008e-01 -1.30999997e-01 1.81970000e+00 -4.51530010e-01 3.95529985e-01 -6.08169973e-01 3.36979985e-01 1.10900000e-01 -2.84759998e-01 3.04610014e-01 -2.16059998e-01 -5.81110008e-02 4.41720009e-01 -2.42310002e-01 -1.27580002e-01 -4.87929992e-02 -1.54949993e-01 -8.54910016e-01 9.29120034e-02 -6.08090013e-02 2.90730000e-02 -3.87349993e-01 -7.08530024e-02 -6.59749985e-01 -3.81570011e-01 5.01699984e-01 -7.35599995e-01 4.15210009e-01 2.13280007e-01 -3.37790012e-01 6.69019997e-01 4.24860001e-01 -1.21480003e-01 -1.06260004e-02 1.27450004e-01 -1.35609999e-01 2.34229997e-01 3.51099998e-01 1.28410006e+00 1.29820004e-01 2.13569999e-01 3.28570008e-01 1.65669993e-01 -2.14579999e-01 -4.42750007e-01 3.28500003e-01 1.80010006e-01 6.48650005e-02 -3.58799994e-01 -1.42259998e-02 3.11250001e-01 -2.20489994e-01 3.28290015e-02 3.85250002e-01 -1.05120003e-01 2.78010011e-01 -1.01709999e-01 -7.15209991e-02] [-0.34636 -0.88984001 -0.50321001 -0.43516001 0.54518998 0.17437001 -0.093541 0.16141 -0.46575999 -0.22 -0.31415001 -0.13484 -0.37617999 -0.67635 0.78820002 -0.33384001 -0.42414001 0.32367 0.50670999 0.21540999 0.43296 1.49049997 0.31795001 -0.15196 0.2579 -0.35666001 -0.63880002 -0.086453 -0.94755 0.19203 0.31161001 -0.74491 -0.59210998 0.4332 -0.064934 -0.48862001 0.35600999 -0.44780999 -0.015773 0.18203001 0.051751 -0.2854 -0.14727999 0.1513 -0.33048001 0.27135 1.16659999 -0.36662 0.090829 0.87301999 -0.13505 0.21204001 0.57270002 0.54259002 -0.50335002 0.16767 -0.82530999 -0.45962 -0.42642 -0.2164 0.088689 -0.15061 -0.16785 -0.31794 -0.69608998 0.40715 -0.29190999 -0.042072 0.90051001 0.35947999 0.030644 -0.028853 0.086821 0.74741 -0.52008998 0.20655 0.44053999 -0.11865 -0.15449999 -0.22457001 -0.15453 0.16101 -0.30825001 -0.28479999 -0.50823998 0.48732999 -0.012029 0.034592 0.48304 -0.56752002 -0.057299 0.22070999 -0.34200001 -0.060634 0.95499998 -0.60952997 0.59577 -0.11553 -0.67475998 0.52658999 0.82163 0.35126001 0.15521 0.12135 0.38191 0.24228001 -0.51485997 1.14810002 0.07281 0.23024 -0.68901998 -0.17606001 -0.24308001 -0.13686 -0.13467 0.059625 -0.68668997 0.15907 0.11983 -0.024954 0.34898001 0.15456 0.047524 0.23616999 0.54784 -1.01380002 0.10351 0.26865 -0.064867 0.23893 0.026141 0.081648 0.74479997 -0.67208999 0.23351 -0.55070001 -0.14296 -0.30303001 -0.40233999 0.012984 0.86865002 -0.14025 1.13900006 -0.093339 1.56060004 0.41804001 0.54067999 -0.43340999 -0.090589 0.56682003 -0.21291 0.45693001 -0.64519 -0.05866 0.21477 0.45563 -0.15218 0.36307001 -0.25441 -0.72013998 0.52191001 0.55509001 -0.073841 0.44994 -0.11501 0.1911 0.077304 0.18629 0.60244 0.028085 0.17228 -0.24455 0.04822 0.51318002 -0.06824 0.35515001 -0.80987 -0.42732999 -0.72728997 0.47817001 0.87611997 -0.18855 0.30390999 -0.14161 0.26699001 -0.81572002 -0.67589998 -0.34687999 0.53188998 0.75443 -0.083874 0.77434999 0.081108 -0.29840001 -0.24409001 -0.14574 -0.1186 0.085964 0.48076999 -0.13097 ] [ 1.07439995 -0.49325001 -0.23126 -0.197 -0.087485 -0.16806 -0.11092 0.42857999 -0.16893999 0.0094633 -0.50453001 -0.40006 0.31169 0.50295001 -0.48537001 -0.20689 -0.62162 0.38407999 0.22182 0.051087 -0.018028 1.37919998 0.3163 -0.17425001 0.11344 -0.42723 -0.28510001 -0.1246 -0.2024 0.18217 -0.37834001 -0.22421999 0.38877001 0.20065001 -0.29708999 0.77051002 0.13087 -0.25084001 0.54755002 0.38086 0.28174999 -0.15691 -0.71987998 0.24118 0.073913 -0.46965 1.02180004 0.049863 0.036841 0.54706001 -0.15903001 0.53780001 -0.10335 0.51644999 -0.25512001 -0.18553001 -0.51804 -0.24337 0.57081997 -0.39017001 -0.17132001 -0.14939 -0.1724 0.91408002 -0.45778 0.40143001 0.075224 -0.4104 -0.1714 -0.63552999 0.60185999 -0.3193 -0.46136999 0.030652 0.32890001 -0.2472 -0.49831 -0.90982997 -0.057251 0.20213 -0.51845998 0.46320999 0.032707 0.29872999 1.11189997 -0.35784999 0.34929001 -0.51739001 0.25235 -1.0309 0.21052 0.06349 -0.10589 0.43222001 0.20389 0.065589 -0.62914002 0.1096 -0.86363 0.44760999 0.43114999 0.041376 -0.42429999 -0.080897 0.093115 0.22603001 0.31176999 0.83004999 -0.25659001 0.013861 0.38326001 -0.52025998 0.30410001 -0.52507001 -0.78566003 -0.046498 0.41154999 -0.21447 -0.24202999 -0.24732 1.01129997 0.067517 0.18331 -0.17636 0.49777001 -0.21067999 0.0037579 0.22881 -0.15993001 -0.13421001 -0.27379999 -0.20734 0.13407999 -0.57762003 -0.66526997 -0.42083001 0.65882999 -0.53825998 -0.50585997 0.51735002 0.25468999 -0.83724999 0.83806998 -0.42219999 1.0776 0.065962 0.48954999 -0.78614002 -0.19338 0.097524 0.27215999 0.037038 0.61778998 -0.29506999 -0.97285002 0.53106999 -0.32765001 -0.045966 -0.75436997 0.024904 0.64872003 0.023095 0.32062 0.35837999 -0.091125 -0.10866 0.33048001 -0.1162 -0.40981999 0.43928999 -0.16706 0.26047 0.090957 0.92714 0.099946 -0.29513001 -0.35341001 0.33693001 -0.42203999 -0.065625 0.54738998 -0.41751 -0.86232001 -0.65891999 -0.41549 0.067035 -0.41558 0.15092 0.17556 0.94068003 0.22527 0.65908003 0.15809 0.061199 0.63612998 -0.17089 -0.017591 -0.054003 -0.69072002 0.65179998] [ 1.03529997e-01 7.20499977e-02 -2.93029994e-02 -4.46799994e-01 -8.61259997e-02 7.40030035e-02 -4.65499997e-01 -6.18570000e-02 -5.03650010e-01 1.95480004e-01 -1.03349999e-01 7.75929987e-01 -1.72040001e-01 -4.53520000e-01 2.63040006e-01 1.64340004e-01 2.80279994e-01 2.89330006e-01 3.10360014e-01 1.27750002e-02 6.79520011e-01 2.62770009e+00 3.89640003e-01 -4.49900001e-01 2.17969999e-01 9.16400030e-02 -1.67940006e-01 7.72420019e-02 2.91269988e-01 1.20530002e-01 -5.49659990e-02 9.16600004e-02 1.31300002e-01 -9.33270007e-02 -3.53009999e-01 -5.03880024e-01 -7.29799986e-01 -3.61380011e-01 -2.99659997e-01 2.07839999e-02 -7.03599975e-02 -6.72269985e-02 -3.62650007e-02 -1.46009997e-02 -5.98580018e-02 -1.63020000e-01 3.00660014e-01 -9.28120017e-02 -6.69799969e-02 1.43830001e-01 1.03950001e-01 -1.93039998e-02 -4.07020003e-01 8.86749983e-01 2.67349988e-01 -1.23829998e-01 2.73739994e-02 3.44639987e-01 6.04049981e-01 2.80640006e-01 1.41320005e-01 2.46429995e-01 -4.48760003e-01 5.42909980e-01 1.96759999e-01 -4.94709998e-01 2.68779993e-02 2.69100010e-01 2.53410012e-01 9.88679975e-02 1.77919999e-01 -3.22290003e-01 -1.05930001e-01 1.89520001e-01 -2.57629991e-01 2.43619993e-01 -2.45560005e-01 -1.36539996e-01 -1.08170003e-01 -4.21220005e-01 -1.61640003e-01 -3.41199994e-01 -1.47389993e-01 -2.16409996e-01 -5.62399998e-02 7.15879977e-01 6.40570000e-02 -3.07790011e-01 6.57369971e-01 -6.36910021e-01 2.64039993e-01 2.15059996e-01 1.83850005e-01 3.32610011e-01 -6.30220026e-02 -2.22450003e-01 6.31980002e-02 -4.47950006e-01 -2.52279997e-01 1.97699994e-01 -3.55479985e-01 1.19139999e-01 -2.99149990e-01 1.29390001e-01 -4.30770010e-01 -1.37700006e-01 2.38429993e-01 1.45529997e+00 -1.13710001e-01 -3.79790008e-01 -2.83129990e-01 -4.64819998e-01 -1.92760006e-01 1.98490005e-02 2.70090014e-01 -1.70849994e-01 -5.20099998e-02 -2.03219995e-01 -3.27439994e-01 -5.07570028e-01 2.98289992e-02 1.80350006e-01 3.05330008e-01 2.40700006e-01 4.66120005e-01 -7.12530017e-01 5.96719980e-01 2.13310003e-01 3.48639995e-01 3.80190015e-01 1.38260007e-01 -7.33380020e-02 2.26349995e-01 -4.55599993e-01 8.19810014e-03 3.47319990e-01 -1.36079997e-01 -6.59820020e-01 -2.79430002e-01 4.81799990e-02 -3.97129990e-02 2.86449999e-01 1.41259998e-01 -3.94299999e-02 1.44519997e+00 4.59950000e-01 7.92850032e-02 -3.55580002e-01 3.09360009e-02 -2.50809994e-02 5.64870015e-02 9.00759995e-02 5.20309992e-02 -3.99529994e-01 3.40330005e-01 -4.17400002e-01 -3.82189989e-01 2.22049996e-01 1.09520003e-01 -1.64539993e-01 1.20609999e-01 1.60919994e-01 -3.27459991e-01 2.45800003e-01 -2.28320006e-02 2.74109989e-01 -2.10689995e-02 3.91460001e-01 -5.61020017e-01 6.95510030e-01 -3.55260000e-02 4.71640006e-02 6.71909988e-01 4.81240004e-01 -2.78840009e-02 5.05490005e-01 -5.41859984e-01 -3.52369994e-01 -3.12009990e-01 -1.76020002e-03 9.59439993e-01 -5.03639996e-01 4.39889997e-01 4.71810013e-01 4.25799996e-01 -5.92209995e-01 -3.96219999e-01 1.89040005e-02 3.33819985e-02 -2.90149987e-01 -1.22079998e-01 2.76309997e-02 -2.66250014e-01 -1.97340008e-02 2.31020004e-01 8.76149982e-02 -7.69149978e-03 1.90050006e-02 -4.42119986e-01 -6.81999996e-02] [ 2.41689995e-01 -3.45340014e-01 -2.23069996e-01 -1.29069996e+00 2.52849996e-01 -5.51280022e-01 -8.03359970e-02 -8.17670021e-03 3.11360002e-01 -4.51009989e-01 2.46610001e-01 3.64410013e-01 9.43359971e-01 -3.54200006e-02 7.80480027e-01 -3.97650003e-01 3.11250001e-01 -1.77430004e-01 -4.19889987e-01 -3.78149986e-01 6.72299981e-01 3.17160010e+00 3.24960016e-02 -3.16400006e-02 5.80680013e-01 -4.44579989e-01 -5.56120016e-02 1.80519998e-01 2.85719991e-01 9.58700031e-02 2.14369997e-01 4.97310013e-02 1.87199995e-01 1.19139999e-01 2.74080001e-02 -8.06079984e-01 -3.08349997e-01 -8.97369981e-01 -1.97720006e-01 2.67409999e-02 -3.87650013e-01 1.16590001e-01 -2.01100007e-01 2.01010004e-01 -7.91329965e-02 -5.09539992e-02 6.01889985e-03 3.34699988e-01 -2.11180001e-01 7.40419999e-02 -2.81410009e-01 -5.96150011e-02 -3.52959991e-01 6.47480011e-01 5.39080016e-02 -3.13760012e-01 -3.66219997e-01 -2.77550012e-01 2.26760004e-02 4.88109998e-02 1.43120006e-01 -1.85800001e-01 -5.69639981e-01 -5.41190028e-01 1.86159998e-01 1.88539997e-01 2.75209993e-01 -1.78350002e-01 -3.74379992e-01 1.21090002e-01 1.86099997e-03 -9.21269972e-03 1.01860002e-01 9.80810001e-02 -3.72449994e-01 6.64150000e-01 5.73659986e-02 -4.38450009e-01 -4.05250013e-01 -5.59589982e-01 -1.13430001e-01 -5.49870014e-01 -2.63209999e-01 -2.84709990e-01 1.44109994e-01 1.03600003e-01 -3.21980000e-01 -2.15299994e-01 9.86679971e-01 -4.19369996e-01 3.11899990e-01 3.33799988e-01 1.60180002e-01 3.31369996e-01 -2.54939999e-02 -3.78799997e-02 -1.20480001e-01 -1.21749997e-01 9.47659984e-02 2.61579990e-01 2.99309995e-02 -2.96249986e-01 4.34009999e-01 -3.65360007e-02 -4.28519994e-01 -3.96380007e-01 -2.49730006e-01 1.10179996e+00 -2.28119999e-01 2.43239999e-01 8.38569999e-02 -4.88810003e-01 -2.13159993e-01 4.02490012e-02 -4.05160010e-01 -1.24109998e-01 -1.97280005e-01 -7.36959994e-01 -4.45820004e-01 -4.49919999e-01 -7.37069994e-02 -1.64539993e-01 1.60109997e-01 -4.19519991e-01 4.14169997e-01 -5.72770000e-01 5.13670027e-01 8.11180025e-02 2.92879995e-02 3.53089988e-01 -1.08029999e-01 1.37020007e-01 3.88280004e-01 -2.88129985e-01 6.82030022e-01 1.74119994e-01 -8.85400027e-02 -2.97850013e-01 -2.84550011e-01 1.35700002e-01 1.52319998e-01 2.20369995e-01 6.88120008e-01 -1.30999997e-01 1.81970000e+00 -4.51530010e-01 3.95529985e-01 -6.08169973e-01 3.36979985e-01 1.10900000e-01 -2.84759998e-01 3.04610014e-01 -2.16059998e-01 -5.81110008e-02 4.41720009e-01 -2.42310002e-01 -1.27580002e-01 -4.87929992e-02 -1.54949993e-01 -8.54910016e-01 9.29120034e-02 -6.08090013e-02 2.90730000e-02 -3.87349993e-01 -7.08530024e-02 -6.59749985e-01 -3.81570011e-01 5.01699984e-01 -7.35599995e-01 4.15210009e-01 2.13280007e-01 -3.37790012e-01 6.69019997e-01 4.24860001e-01 -1.21480003e-01 -1.06260004e-02 1.27450004e-01 -1.35609999e-01 2.34229997e-01 3.51099998e-01 1.28410006e+00 1.29820004e-01 2.13569999e-01 3.28570008e-01 1.65669993e-01 -2.14579999e-01 -4.42750007e-01 3.28500003e-01 1.80010006e-01 6.48650005e-02 -3.58799994e-01 -1.42259998e-02 3.11250001e-01 -2.20489994e-01 3.28290015e-02 3.85250002e-01 -1.05120003e-01 2.78010011e-01 -1.01709999e-01 -7.15209991e-02] [ 0.041052 -0.54705 -0.72193998 -0.31235999 -0.43849 0.10691 -0.50641 -0.45401001 -0.28623 0.018973 0.020495 0.42860001 -0.0057162 -0.21272001 0.71864998 -0.091906 -0.55365002 0.39133 0.15351 -0.27454001 0.56528002 3.04830003 0.30467001 -0.37893 0.37865999 0.13925999 -0.11482 0.48212999 -0.30522999 0.43125001 -0.09667 0.069156 0.31426001 0.26350999 -0.31189999 -0.39881 -0.55656999 -0.35934001 -0.25402001 0.072061 -0.12966999 -0.11247 -0.041192 -0.042619 -0.07848 0.31992 0.41635999 0.26131001 -0.18175 -0.1279 0.21332 -0.41973001 -0.50444001 0.37705001 0.83955002 -0.34571001 -0.43000999 -0.18653999 -0.061082 -0.087612 0.092833 0.52604997 -0.57611001 -0.19328 0.20576 0.24607 0.1631 -0.18138 0.032592 0.19169 0.73565 -0.25718999 0.30072999 0.56699002 -0.21544001 0.18933 -0.12287 -0.65759999 0.021702 0.1041 0.098952 -0.43171999 -0.27517 -0.15448 0.31301001 -0.032041 -0.090526 0.14489999 0.68151999 -0.88242 0.30816999 -0.62702 0.12274 0.014773 -0.16887 0.56159002 0.022004 0.52085 -0.22685 0.09633 0.26956999 0.30489001 0.018463 0.31009001 0.04198 0.32381999 0.13153 0.89722002 0.15475 -0.38806999 -0.52993 -0.35383999 -0.0913 0.57230002 0.48067001 0.24438 0.074138 -0.019957 -0.35030001 -0.034695 -0.12045 0.39998999 -0.37015 -0.53040999 0.10655 -0.44973001 0.43105 -0.44937 0.48675999 0.43836999 0.043421 0.52675003 0.61176002 0.26704001 0.59239 0.23650999 0.12841 -0.10665 -0.46715 -0.039081 -0.24921 0.030486 0.092933 -0.04841 1.83580005 0.077535 -0.11494 -0.13668001 -0.23983 0.31948 0.19205999 0.38894001 0.34755 -0.038804 0.19541 -0.37099999 -0.027576 -0.24127001 -0.16868 0.032815 0.08139 0.054121 -0.42785999 0.26447001 0.054847 -0.21765999 0.015181 0.57656002 0.24033 0.62076002 -0.019055 -0.31509 0.76424998 0.35168999 -0.28512001 0.15175 0.11238 -0.60829997 0.35087001 0.19140001 0.51753998 0.20893 -0.63814002 0.19403 0.24493 0.46606001 -0.32235 0.37286001 -0.19508 0.13237999 -0.35420999 0.22849 0.36032 -0.0050241 -0.051955 -0.37755999 -0.087065 0.3592 0.11564 0.44372001] [ 4.71520007e-01 -5.87790012e-01 -6.76150024e-01 -4.67000008e-01 1.11709997e-01 3.26370001e-01 -4.65070009e-01 -8.05180013e-01 -1.65340006e-01 -1.13849998e-01 -1.36849999e-01 2.56980002e-01 3.36279988e-01 1.90149993e-01 1.32440001e-01 2.37419993e-01 -1.46239996e+00 8.59059989e-01 5.53650022e-01 1.94330007e-01 2.73930013e-01 1.05669999e+00 6.24970019e-01 -4.30469990e-01 7.14770019e-01 -4.73030001e-01 -8.97960007e-01 2.56910007e-02 -6.42499983e-01 2.15990007e-01 -1.22769997e-01 -5.36949992e-01 5.91489971e-01 6.28649965e-02 1.51260002e-02 -6.57150000e-02 1.61709994e-01 -8.86740014e-02 -7.09370002e-02 6.12349987e-01 1.38689995e-01 -3.67980003e-01 -9.46219981e-01 1.30669996e-01 -2.82139987e-01 -3.02709997e-01 4.05889988e-01 -2.11899996e-01 1.74940005e-01 2.38450006e-01 3.41769993e-01 4.50269997e-01 -7.82140017e-01 1.64210007e-01 7.19319999e-01 -6.80140018e-01 -4.93660003e-01 3.67380008e-02 2.62410015e-01 -8.48299980e-01 -6.59759998e-01 4.04370010e-01 -2.61209998e-02 5.83829999e-01 -3.28000009e-01 6.39530003e-01 1.20350003e-01 7.21519988e-04 8.28130007e-01 -3.83879989e-01 5.35929978e-01 -4.59630013e-01 -5.12839973e-01 1.74339995e-01 2.06220001e-01 -8.01329970e-01 -4.74339992e-01 -4.32810009e-01 -6.11400008e-01 1.71409994e-01 5.54369986e-01 6.11240007e-02 6.43959999e-01 -5.23599982e-01 1.35130000e+00 -1.40279993e-01 3.67210001e-01 -3.68629992e-01 7.95690000e-01 -1.01139998e+00 -1.47060007e-01 4.48889993e-02 4.06240001e-02 5.33150017e-01 3.40680003e-01 2.50710011e-01 -1.26489997e-01 -1.15050003e-01 -1.48660004e-01 7.65860021e-01 9.80419964e-02 6.28759980e-01 -4.09599990e-01 -3.33020017e-02 -3.77389997e-01 -2.54250001e-02 -8.92150030e-02 1.63540006e+00 5.04270017e-01 -3.86139989e-01 -1.25259995e-01 1.65910006e-01 -2.19990000e-01 -6.84350014e-01 -2.99389988e-01 -1.25550002e-01 3.63970011e-01 3.89310002e-01 -9.63360012e-01 -7.42670000e-02 3.48879993e-01 2.36190006e-01 -8.27549994e-01 3.19779992e-01 -3.00359994e-01 -4.01740015e-01 7.04670012e-01 -3.32989991e-01 -1.26369998e-01 -3.72830003e-01 -9.09730017e-01 -1.33279994e-01 -5.32779992e-02 -3.47909987e-01 2.48980001e-01 -4.34430003e-01 -2.42050007e-01 -6.21890008e-01 -1.27769995e+00 -9.66310024e-01 4.20859993e-01 -4.12339985e-01 1.20589995e+00 -2.55379993e-02 1.01170003e+00 -1.11219997e-03 5.92599988e-01 -5.58459997e-01 -3.69489998e-01 5.58549985e-02 -6.60969973e-01 3.91559988e-01 3.09260011e-01 -6.28539994e-02 -1.14900005e+00 4.69440013e-01 -5.05439997e-01 -1.75500005e-01 -3.80259991e-01 -1.97349995e-01 7.31180012e-01 -1.77149996e-01 4.18940008e-01 4.23940003e-01 -5.16149998e-02 -1.20180003e-01 3.96189988e-01 -2.36780003e-01 1.19029999e-01 2.93020010e-01 -1.81150004e-01 4.59089994e-01 1.93489999e-01 4.75919992e-01 -3.67170006e-01 -8.76540027e-04 -6.15379997e-02 -4.12099995e-02 -1.95240006e-01 -6.47260016e-03 4.30709988e-01 6.84989989e-02 -4.27579999e-01 -3.15990001e-01 -2.59119987e-01 -6.58209980e-01 -1.92790002e-01 5.57280004e-01 -2.24089995e-01 2.21340001e-01 4.68760014e-01 6.37290001e-01 6.58949971e-01 8.79120007e-02 -3.06369990e-01 -3.22079986e-01 6.47899985e-01 1.75490007e-01 -5.78580022e-01 8.94800007e-01]
===================================
===== For the word2vec
===================================
1. Download and unzip the pre-trained model of GoogleNews-vectors-negative300.bin.gz.
2. Install the gensim tools:
sudo pip install --upgrade gensim
3. Code for vector extraction from given sentence.
import gensim
print("==>> loading the pre-trained word2vec model: GoogleNews-vectors-negative300.bin")
dictFileName = './GoogleNews-vectors-negative300.bin'
wv = gensim.models.KeyedVectors.load_word2vec_format(dictFileName, binary=True)
The Output is:
==>> loading the pre-trained word2vec model: GoogleNews-vectors-negative300.bin
INFO:gensim.models.utils_any2vec:loading projection weights from ./GoogleNews-vectors-negative300.bin
INFO:gensim.models.utils_any2vec:loaded (3000000, 300) matrix from ./GoogleNews-vectors-negative300.bin
INFO:root:Data statistic
INFO:root:train_labels:19337
INFO:root:test_labels:20632
INFO:root:train_sentences:19337
INFO:root:dev_sentences:2000
INFO:root:test_sentences:20632
INFO:root:dev_labels:2000
embed_size:300
vocab_size:3000000
vocab_path = ''data/bilstm.vocab''
index_to_word = [key for key in wv.vocab]
word_to_index = {}
for index, word in enumerate(index_to_word):
word_to_index[word] = index
with open(vocab_path, "w") as f:
f.write(json.dumps(word_to_index))
Let's take a deep understanding on the bidirectional-LSTM-for-text-classification-master
1 class BiLSTM(nn.Module): 2 def __init__(self, embedding_matrix, hidden_size=150, num_layer=2, embedding_freeze=False): 3 super(BiLSTM,self).__init__() 4 5 # embedding layer 6 vocab_size = embedding_matrix.shape[0] 7 embed_size = embedding_matrix.shape[1] 8 self.hidden_size = hidden_size 9 self.num_layer = num_layer 10 self.embed = nn.Embedding(vocab_size, embed_size) 11 self.embed.weight = nn.Parameter(torch.from_numpy(embedding_matrix).type(torch.FloatTensor), requires_grad=not embedding_freeze) 12 self.embed_dropout = nn.Dropout(p=0.3) 13 self.custom_params = [] 14 if embedding_freeze == False: 15 self.custom_params.append(self.embed.weight) 16 17 # The first LSTM layer 18 self.lstm1 = nn.LSTM(embed_size, self.hidden_size, num_layer, dropout=0.3, bidirectional=True) 19 for param in self.lstm1.parameters(): 20 self.custom_params.append(param) 21 if param.data.dim() > 1: 22 nn.init.orthogonal(param) 23 else: 24 nn.init.normal(param) 25 26 self.lstm1_dropout = nn.Dropout(p=0.3) 27 28 # The second LSTM layer 29 self.lstm2 = nn.LSTM(2*self.hidden_size, self.hidden_size, num_layer, dropout=0.3, bidirectional=True) 30 for param in self.lstm2.parameters(): 31 self.custom_params.append(param) 32 if param.data.dim() > 1: 33 nn.init.orthogonal(param) 34 else: 35 nn.init.normal(param) 36 self.lstm2_dropout = nn.Dropout(p=0.3) 37 38 # Attention 39 self.attention = nn.Linear(2*self.hidden_size,1) 40 self.attention_dropout = nn.Dropout(p=0.5) 41 42 # Fully-connected layer 43 self.fc = weight_norm(nn.Linear(2*self.hidden_size,3)) 44 for param in self.fc.parameters(): 45 self.custom_params.append(param) 46 if param.data.dim() > 1: 47 nn.init.orthogonal(param) 48 else: 49 nn.init.normal(param) 50 51 self.hidden1=self.init_hidden() 52 self.hidden2=self.init_hidden() 53 54 def init_hidden(self, batch_size=3): 55 if torch.cuda.is_available(): 56 return (Variable(torch.zeros(self.num_layer*2, batch_size, self.hidden_size)).cuda(), 57 Variable(torch.zeros(self.num_layer*2, batch_size, self.hidden_size)).cuda()) 58 else: 59 return (Variable(torch.zeros(self.num_layer*2, batch_size, self.hidden_size)), 60 Variable(torch.zeros(self.num_layer*2, batch_size, self.hidden_size))) 61 62 def forward(self, sentences): 63 64 print("==>> sentences: ", sentences) 65 66 # get embedding vectors of input 67 padded_sentences, lengths = torch.nn.utils.rnn.pad_packed_sequence(sentences, padding_value=int(0), batch_first=True) 68 print("==>> padded_sentences: ", padded_sentences) 69 70 71 embeds = self.embed(padded_sentences) 72 print("==>> embeds: ", embeds) 73 74 # pdb.set_trace() 75 76 noise = Variable(torch.zeros(embeds.shape).cuda()) 77 noise.data.normal_(std=0.3) 78 embeds += noise 79 embeds = self.embed_dropout(embeds) 80 # add noise 81 82 packed_embeds = torch.nn.utils.rnn.pack_padded_sequence(embeds, lengths, batch_first=True) 83 84 print("==>> packed_embeds: ", packed_embeds) 85 86 87 88 # First LSTM layer 89 # self.hidden = num_layers*num_directions batch_size hidden_size 90 packed_out_lstm1, self.hidden1 = self.lstm1(packed_embeds, self.hidden1) 91 padded_out_lstm1, lengths = torch.nn.utils.rnn.pad_packed_sequence(packed_out_lstm1, padding_value=int(0)) 92 padded_out_lstm1 = self.lstm1_dropout(padded_out_lstm1) 93 packed_out_lstm1 = torch.nn.utils.rnn.pack_padded_sequence(padded_out_lstm1, lengths) 94 95 96 97 pdb.set_trace() 98 99 # Second LSTM layer 100 packed_out_lstm2, self.hidden2 = self.lstm2(packed_out_lstm1, self.hidden2) 101 padded_out_lstm2, lengths = torch.nn.utils.rnn.pad_packed_sequence(packed_out_lstm2, padding_value=int(0), batch_first=True) 102 padded_out_lstm2 = self.lstm2_dropout(padded_out_lstm2) 103 104 # attention 105 unnormalize_weight = F.tanh(torch.squeeze(self.attention(padded_out_lstm2), 2)) 106 unnormalize_weight = F.softmax(unnormalize_weight, dim=1) 107 unnormalize_weight = torch.nn.utils.rnn.pack_padded_sequence(unnormalize_weight, lengths, batch_first=True) 108 unnormalize_weight, lengths = torch.nn.utils.rnn.pad_packed_sequence(unnormalize_weight, padding_value=0.0, batch_first=True) 109 logging.debug("unnormalize_weight size: %s" % (str(unnormalize_weight.size()))) 110 normalize_weight = torch.nn.functional.normalize(unnormalize_weight, p=1, dim=1) 111 normalize_weight = normalize_weight.view(normalize_weight.size(0), 1, -1) 112 weighted_sum = torch.squeeze(normalize_weight.bmm(padded_out_lstm2), 1) 113 114 # fully connected layer 115 output = self.fc(self.attention_dropout(weighted_sum)) 116 return output
==>> Some Testing:
(a). len(wv.vocab) = 300,0000
(b). wv.vocab is what ? Something like this:
{ ... , u'fivemonth': <gensim.models.keyedvectors.Vocab object at 0x7f90945bd810>,
u'retractable_roofs_Indians': <gensim.models.keyedvectors.Vocab object at 0x7f90785f5690>,
u'Dac_Lac_province': <gensim.models.keyedvectors.Vocab object at 0x7f908d8eda10>,
u'Kenneth_Klinge': <gensim.models.keyedvectors.Vocab object at 0x7f9081563410>}
(c). index_to_word: count the words from the pre-trained model.
{ ... , u"Lina'la_Sin_Casino", u'fivemonth', u'retractable_roofs_Indians', u'Dac_Lac_province', u'Kenneth_Klinge'}
(d). word_to_index: give each word a index as following
{ ... , u'fivemonth': 2999996, u'Pidduck': 2999978, u'Dac_Lac_province': 2999998, u'Kenneth_Klinge': 2999999 }
(e). dataset is:
[ ... , [1837614, 1569052, 1837614, 1288695, 2221039, 2323218, 1837614, 1837614, 2029395, 1612781, 311032, 1524921, 1837614, 2973515, 2033866, 882731, 2462275, 2809106, 1479961, 826019, 73590, 953550, 1837614],
[1524921, 1113778, 1837614, 318169, 1837614, 1954969, 196613, 943118, 1837614, 2687790, 291413, 2774825, 2366038, 296869, 1468080, 856987, 1802099, 724308, 1207907, 2264894, 2206446, 812434],
[564298, 477983, 1837614, 1449153, 1837614, 211925, 2206446, 481834, 488597, 280760, 2072822, 1344872, 1791678, 2458776, 2965810, 2168205, 387112, 2656471, 1391, 1837614, 1801696, 2093846, 1210651],
[2493381, 133883, 2441902, 1014220, 1837614, 2597880, 1756105, 2651537, 1391, 2641114, 2517536, 1109601, 122269, 1782479, 2965835, 488597, 1767716, 753333, 564298, 2380935, 228060, 1837614, 371618],
[1837614, 1344872, 2458776, 2965810, 1837614, 2015408, 1837614, 528014, 1991322, 1837614, 908982, 1484130, 2349526, 988689, 753336, 1837614, 364492, 2183116, 826019, 73590, 953550, 1837614],
[1837614, 2673785, 1990947, 1219831, 2635341, 1247040, 1837614, 799543, 1990947, 1219831, 2722301, 1837614, 1427513, 969099, 2157673, 1430111, 1837614]]
(f). the variable during the training process:
# ('==>> sentences: ', PackedSequence(data=tensor(
# [ 4.0230e+05, 1.8376e+06, 2.0185e+06, 1.8376e+06, 2.8157e+06,
# 1.8376e+06, 1.8376e+06, 1.0394e+06, 1.8376e+06, 2.9841e+06,
# 4.4713e+05, 1.1352e+06, 2.3532e+06, 1.8376e+06, 1.8376e+06,
# 1.8376e+06, 1.9550e+06, 3.8429e+04, 6.2537e+05, 2.3764e+05,
# 1.8376e+06, 1.5428e+06, 1.4214e+06], device='cuda:0'),
# batch_sizes=tensor([ 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1])))
# ('==>> padded_sentences: ', tensor(
# [[ 4.0230e+05, 1.8376e+06, 2.0185e+06, 1.8376e+06, 2.8157e+06,
# 1.8376e+06, 1.8376e+06, 1.0394e+06, 1.8376e+06, 2.9841e+06,
# 4.4713e+05, 1.1352e+06, 2.3532e+06, 1.8376e+06, 1.8376e+06,
# 1.8376e+06, 1.9550e+06, 3.8429e+04, 6.2537e+05, 2.3764e+05,
# 1.8376e+06, 1.5428e+06, 1.4214e+06]], device='cuda:0'))
# length: 23
# ('==>> embeds: ', tensor([[
# [-0.0684, 0.1826, -0.1777, ..., 0.1904, -0.1021, 0.1729],
# [ 0.0801, 0.1050, 0.0498, ..., 0.0037, 0.0476, -0.0688],
# [-0.1982, -0.0693, 0.1230, ..., -0.1357, -0.0306, 0.1104],
# ...,
# [ 0.0801, 0.1050, 0.0498, ..., 0.0037, 0.0476, -0.0688],
# [-0.0518, -0.0299, 0.0415, ..., 0.0776, -0.1660, 0.1602],
# [-0.0532, -0.0004, 0.0337, ..., -0.2373, -0.1709, 0.0233]]], device='cuda:0'))
# ('==>> packed_embeds: ', PackedSequence(data=tensor([
# [-0.3647, 0.2966, -0.2359, ..., -0.0000, 0.2657, -0.4302],
# [ 1.1699, 0.0000, 0.3312, ..., 0.5714, 0.1930, -0.2267],
# [-0.0627, -1.0548, 0.4966, ..., -0.5135, -0.0150, -0.0000],
# ...,
# [-0.6065, -0.7562, 0.3320, ..., -0.5854, -0.2089, -0.5737],
# [-0.0000, 0.4390, 0.0000, ..., 0.6891, 0.0250, -0.0000],
# [ 0.6909, -0.0000, -0.0000, ..., 0.1867, 0.0594, -0.2385]], device='cuda:0'),
# batch_sizes=tensor([ 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1])))
# packed_out_lstm1
# PackedSequence(data=tensor([
# [-0.0000, -0.0000, 0.1574, ..., 0.1864, -0.0901, -0.3205],
# [-0.0000, -0.3490, 0.1774, ..., 0.1677, -0.0000, -0.3688],
# [-0.3055, -0.0000, 0.2240, ..., 0.0000, -0.0927, -0.0000],
# ...,
# [-0.3188, -0.4134, 0.1339, ..., 0.3161, -0.0000, -0.3846],
# [-0.3355, -0.4365, 0.1575, ..., 0.2775, -0.0886, -0.4015],
# [-0.0000, -0.0000, 0.2452, ..., 0.1763, -0.0000, -0.2748]], device='cuda:0'),
# batch_sizes=tensor([ 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1]))
## self.hidden1
1 # self.hidden1 2 # (tensor([ 3 # [[-0.0550, -0.4266, 0.0498, 0.0750, 0.3122, 0.0499, -0.2899, 4 # -0.0980, -0.1776, 0.4376, 0.1555, 0.7167, -0.0162, 0.0081, 5 # -0.3512, 0.0250, 0.0948, -0.3596, 0.0670, -0.6623, 0.0026, 6 # -0.0262, 0.1934, 0.3206, -0.1941, 0.3229, -0.6488, -0.3858, 7 # 0.3169, -0.0216, -0.1136, 0.1411, -0.0296, -0.2860, -0.1277, 8 # -0.0154, 0.2620, -0.3553, 0.0637, 0.8245, 0.6886, -0.0048, 9 # -0.1717, 0.2495, -0.8636, -0.0217, -0.2365, 0.1324, 0.1790, 10 # -0.1515, 0.3530, -0.1644, 0.0073, 0.6709, 0.1577, -0.0190, 11 # 0.0384, 0.4871, -0.7375, -0.1804, 0.3034, -0.3516, -0.2870, 12 # 0.6387, 0.0414, 0.6983, -0.3211, 0.0449, 0.2127, -0.0421, 13 # -0.3454, -0.8145, -0.3629, -0.0828, -0.1558, 0.4048, -0.4971, 14 # -0.7495, 0.0622, -0.3318, 0.3913, -0.0322, -0.0678, 0.0307, 15 # -0.0153, -0.1535, -0.2719, 0.0128, -0.1521, -0.2924, 0.7109, 16 # -0.8551, 0.0330, 0.0482, 0.2410, -0.0655, -0.2496, 0.1816, 17 # 0.4963, -0.7593, -0.0022, -0.1122, 0.6857, -0.5693, 0.5805, 18 # 0.7660, 0.4430, 0.0243, -0.0313, 0.0780, -0.2419, 0.0745, 19 # -0.0119, 0.5761, -0.0285, -0.1085, -0.1783, -0.0706, 0.1243, 20 # 0.6333, 0.0296, 0.5557, 0.2717, -0.0071, 0.0503, -0.0405, 21 # -0.4542, 0.8905, 0.4492, 0.8809, 0.7021, 0.8616, 0.2825, 22 # -0.2114, 0.3026, -0.1384, 0.1252, 0.4989, 0.2236, -0.5374, 23 # -0.1352, -0.0561, 0.0378, -0.5291, -0.1004, -0.3723, 0.0836, 24 # 0.3500, 0.0542, 0.2013]], 25 26 # [[-0.0041, -0.0040, -0.1428, 0.2783, -0.1378, -0.4242, 0.1000, 27 # 0.1641, -0.0175, -0.0896, 0.3241, 0.3513, -0.4675, -0.1250, 28 # 0.0546, 0.2400, -0.0997, -0.5614, 0.2026, -0.1505, 0.0833, 29 # -0.3128, -0.4646, -0.0778, 0.2204, 0.5597, -0.7004, 0.0419, 30 # -0.3699, -0.5748, 0.7741, -0.7220, 0.0494, 0.3430, 0.1389, 31 # 0.0178, 0.0136, 0.0273, 0.1559, 0.4333, 0.2411, 0.0804, 32 # -0.0202, 0.6204, -0.4104, 0.5382, 0.1804, 0.0179, 0.1118, 33 # 0.3084, 0.1894, 0.0092, -0.2182, 0.0022, 0.4377, -0.0575, 34 # 0.0906, -0.0531, 0.0936, -0.1203, -0.0836, -0.1718, 0.2059, 35 # 0.7192, 0.0833, 0.3712, 0.2354, -0.1141, -0.0993, -0.0433, 36 # -0.1474, -0.0078, 0.0834, 0.0864, 0.9305, -0.3387, 0.0818, 37 # 0.3775, 0.1609, -0.0246, 0.2563, -0.1253, -0.0897, 0.0322, 38 # -0.0848, 0.0673, -0.1051, 0.0205, 0.0183, -0.0007, 0.1229, 39 # -0.3388, -0.0948, 0.0335, 0.0450, -0.2747, 0.2763, -0.2691, 40 # 0.6240, -0.0018, 0.2048, 0.3943, 0.2015, -0.5962, -0.0069, 41 # -0.3460, -0.7910, 0.3002, -0.4653, -0.0611, 0.6912, -0.8154, 42 # -0.0443, -0.0189, -0.1265, -0.1202, 0.0013, -0.5983, 0.0879, 43 # 0.0752, 0.8593, 0.0357, 0.0953, -0.0525, 0.2069, -0.6292, 44 # -0.0456, -0.7646, 0.0166, 0.0584, 0.0142, 0.0575, -0.1658, 45 # -0.4304, 0.3228, 0.4094, 0.0149, 0.1478, 0.1447, 0.4192, 46 # -0.0783, -0.0683, 0.0259, 0.0665, 0.6224, 0.3775, 0.0247, 47 # 0.1710, -0.3622, 0.0931]], 48 49 # [[-0.2315, -0.2179, 0.1716, -0.6143, 0.4329, 0.2288, 0.1208, 50 # -0.3435, 0.6314, 0.0106, 0.1470, -0.0915, -0.3019, 0.1302, 51 # 0.2325, 0.1794, 0.0145, -0.6598, 0.0062, -0.0743, 0.0232, 52 # 0.9310, -0.8155, -0.0449, -0.5504, 0.5746, 0.3607, 0.4053, 53 # -0.1887, -0.5448, 0.1522, -0.0189, -0.4852, 0.6322, 0.0011, 54 # -0.1590, -0.0054, 0.2972, -0.0270, 0.0047, 0.2944, -0.0629, 55 # -0.1138, 0.1349, 0.5449, 0.8018, -0.0218, 0.0523, 0.3262, 56 # -0.0506, 0.2821, -0.0661, -0.7165, -0.3653, 0.3321, -0.0255, 57 # -0.0551, -0.1826, 0.6027, -0.1995, 0.0598, 0.0205, -0.1769, 58 # -0.0789, 0.4500, -0.1641, 0.5002, -0.0716, -0.3708, -0.0020, 59 # -0.5195, -0.0896, 0.1421, 0.1149, 0.3407, 0.2649, 0.0858, 60 # 0.2778, -0.3768, 0.6176, -0.2148, 0.5444, 0.3009, 0.4848, 61 # -0.1174, 0.0019, 0.6213, 0.2524, -0.0816, 0.4639, 0.4747, 62 # -0.7812, 0.2435, -0.0867, 0.1617, 0.2194, 0.0426, 0.1393, 63 # -0.0448, 0.0506, -0.5524, 0.0707, 0.2226, -0.0337, 0.7445, 64 # -0.4516, 0.1107, -0.2617, -0.1914, 0.7238, -0.2689, 0.0110, 65 # -0.3139, -0.0027, -0.5964, -0.9012, -0.4319, 0.0112, -0.0306, 66 # 0.4002, -0.1117, -0.1021, 0.1652, -0.2872, 0.3640, 0.2162, 67 # -0.3843, -0.0869, -0.1623, 0.0297, -0.0048, -0.0735, -0.0886, 68 # -0.4138, 0.2325, -0.4248, 0.3354, 0.0712, -0.4079, 0.0821, 69 # 0.1413, 0.2241, -0.1938, -0.0807, 0.3551, -0.0814, 0.1438, 70 # -0.6870, -0.3647, 0.0276]], 71 72 # [[ 0.0258, 0.3281, -0.8145, -0.0476, -0.2886, -0.8013, 0.2135, 73 # 0.1541, 0.2069, 0.1345, -0.0171, -0.0228, -0.5237, 0.4917, 74 # 0.5187, 0.1402, 0.0928, -0.0373, 0.2698, 0.1259, 0.0021, 75 # -0.1624, -0.4100, 0.5377, -0.1013, -0.5658, -0.1015, 0.5609, 76 # 0.1661, 0.5731, 0.0012, -0.1766, 0.0743, -0.3630, 0.1082, 77 # 0.4643, 0.0175, -0.0260, 0.3810, -0.6425, -0.5515, 0.8800, 78 # -0.1158, -0.5741, 0.0463, 0.4033, 0.0803, 0.0403, 0.1159, 79 # 0.4471, 0.0294, 0.2899, 0.0248, -0.1772, 0.6600, -0.2252, 80 # -0.4896, -0.1285, -0.2377, -0.4179, -0.4056, -0.3224, -0.6855, 81 # -0.2703, 0.2971, 0.1259, -0.0456, -0.2495, 0.8141, 0.4453, 82 # 0.7480, -0.0578, 0.8023, -0.3586, -0.5229, 0.2299, 0.9668, 83 # -0.0717, 0.5355, -0.0743, 0.5246, 0.1604, -0.1464, -0.0757, 84 # 0.0414, -0.0861, 0.2245, 0.1247, 0.0676, -0.2053, 0.0113, 85 # 0.7875, -0.0308, 0.2025, 0.1289, -0.0020, -0.3099, 0.5317, 86 # -0.0117, 0.0928, -0.4100, -0.6184, 0.1171, 0.0216, -0.1266, 87 # 0.1640, 0.0821, -0.4097, -0.0691, 0.5805, 0.1692, -0.2021, 88 # 0.5971, 0.1172, -0.6535, -0.0579, 0.1177, 0.1123, -0.1943, 89 # 0.0488, -0.1305, -0.4859, -0.2758, -0.2972, -0.0605, -0.0029, 90 # -0.1508, 0.0375, -0.5942, -0.2139, -0.0335, -0.2320, -0.1152, 91 # -0.2054, -0.2643, -0.1770, 0.1245, 0.6334, -0.0363, 0.0264, 92 # -0.3348, -0.0434, -0.3794, -0.0913, 0.1293, -0.6537, 0.6490, 93 # 0.1305, -0.0631, -0.2243]]], device='cuda:0'), 94 # tensor([[[ -0.0775, -4.6346, 7.9077, 0.1164, 0.8626, 0.4240, 95 # -0.9286, -0.1612, -0.6049, 0.6771, 0.7443, 1.7457, 96 # -0.3930, 0.0112, -13.5393, 0.0317, 0.1236, -0.8475, 97 # 0.1212, -1.3623, 0.0117, -0.3297, 0.9009, 0.3840, 98 # -0.5885, 0.7411, -8.9613, -1.1402, 0.4511, -0.0753, 99 # -0.3107, 1.6518, -0.0870, -3.1360, -12.0200, -0.0464, 100 # 0.2756, -0.6695, 0.1604, 4.8299, 6.4623, -0.9555, 101 # -0.6904, 0.4469, -11.3343, -0.1669, -0.2747, 0.1590, 102 # 0.5829, -0.3345, 2.1731, -0.5636, 0.0207, 0.9874, 103 # 0.6291, -1.2261, 0.1946, 1.1287, -1.5759, -0.1875, 104 # 0.5550, -1.7350, -0.8235, 1.5122, 0.2019, 5.5143, 105 # -3.8153, 0.6771, 0.3011, -0.2994, -0.7320, -1.5857, 106 # -0.4785, -0.5584, -0.3226, 1.1932, -4.3901, -1.6923, 107 # 0.3526, -1.0625, 0.9279, -0.1843, -0.4376, 2.2389, 108 # -0.1558, -0.3959, -1.2987, 0.0279, -0.4938, -0.3364, 109 # 3.2596, -2.2647, 0.1448, 0.0726, 0.3968, -0.1885, 110 # -0.3960, 0.2141, 0.6785, -2.1622, -0.0043, -0.7516, 111 # 0.9367, -0.7092, 4.1853, 1.3348, 0.6993, 0.2043, 112 # -0.0916, 0.1392, -1.5672, 0.0867, -0.0346, 0.9226, 113 # -0.0470, -0.6870, -0.3002, -0.1131, 0.7785, 1.0582, 114 # 0.0914, 2.8785, 0.8164, -0.1048, 0.0573, -0.0499, 115 # -0.5990, 3.1714, 0.7925, 1.7461, 1.3243, 2.9236, 116 # 0.8966, -0.4455, 0.8763, -0.3036, 0.3302, 2.6581, 117 # 0.4608, -0.7280, -2.9457, -0.1973, 0.0585, -0.6555, 118 # -0.6621, -0.4549, 0.5812, 0.4495, 0.1350, 1.8521]], 119 120 # [[ -0.0053, -0.0843, -0.1763, 0.3929, -0.1668, -0.6609, 121 # 0.1269, 0.2214, -0.0208, -0.3571, 0.7532, 0.7496, 122 # -4.9288, -0.5457, 0.3557, 0.4795, -0.2318, -0.9659, 123 # 0.6826, -1.6542, 0.4917, -0.3956, -1.5164, -0.2274, 124 # 0.6779, 1.1201, -3.1397, 0.0434, -0.4993, -0.8809, 125 # 6.1257, -5.6283, 0.4273, 1.5070, 0.6624, 0.1289, 126 # 0.2180, 0.9920, 0.1646, 0.8828, 0.5732, 0.3255, 127 # -0.0679, 0.9843, -1.8408, 1.0547, 0.1959, 0.0748, 128 # 0.1907, 0.4751, 0.3174, 0.0747, -0.6487, 0.0377, 129 # 0.5554, -0.4095, 0.2593, -0.0568, 0.3751, -0.3646, 130 # -0.2031, -0.3284, 0.4058, 1.2788, 0.1348, 1.8184, 131 # 0.8482, -0.7494, -0.2395, -0.4352, -0.1584, -0.0105, 132 # 0.2676, 0.3763, 2.1413, -1.3001, 0.3923, 1.6432, 133 # 0.2987, -1.2708, 5.5667, -0.1727, -0.5106, 0.5180, 134 # -6.7258, 0.5001, -0.3052, 0.0843, 0.0474, -0.2306, 135 # 0.1908, -2.2523, -1.5432, 0.2809, 0.7099, -0.4145, 136 # 0.7393, -0.6529, 0.7825, -0.0019, 1.3337, 2.2042, 137 # 9.2887, -0.8515, -0.0610, -0.6146, -1.5616, 0.3592, 138 # -1.3585, -0.2641, 1.4763, -3.2525, -0.5447, -0.0453, 139 # -1.0416, -2.4657, 0.0556, -0.7654, 0.2062, 0.0855, 140 # 3.0740, 0.0952, 0.4923, -0.1772, 0.9173, -4.2004, 141 # -0.1298, -2.4266, 0.0181, 0.5039, 0.0399, 7.9909, 142 # -0.5778, -2.9112, 0.4854, 1.2364, 0.0686, 0.6365, 143 # 0.1869, 0.6050, -0.1246, -0.1848, 0.5406, 0.2110, 144 # 1.2367, 1.9466, 0.0302, 0.2002, -0.5902, 0.1069]], 145 146 # [[ -0.3559, -0.2859, 0.2699, -1.4359, 0.9814, 0.2811, 147 # 0.8539, -2.6654, 2.5455, 0.0434, 0.5947, -0.5325, 148 # -0.4638, 0.5487, 1.4376, 0.9863, 0.7429, -1.8308, 149 # 0.0402, -0.2282, 0.0366, 12.7877, -1.2491, -0.1437, 150 # -1.4960, 0.7364, 0.8599, 1.8343, -3.5117, -1.2758, 151 # 0.2930, -0.0472, -0.7527, 0.9555, 0.0446, -0.3389, 152 # -0.1985, 1.7953, -0.5702, 0.0141, 0.6166, -0.0924, 153 # -0.5182, 0.5146, 2.0801, 2.7460, -0.2606, 0.2090, 154 # 0.9266, -0.4758, 0.9961, -0.1723, -1.2069, -1.1735, 155 # 0.3683, -0.4933, -0.0604, -0.2354, 0.8239, -5.4226, 156 # 0.0854, 0.1185, -0.2656, -0.2689, 0.6047, -0.6246, 157 # 1.0131, -0.1673, -0.4990, -0.0690, -0.6092, -0.5205, 158 # 0.1808, 0.3061, 0.3924, 0.5868, 0.1452, 2.8930, 159 # -0.6085, 1.6086, -0.4763, 5.0389, 1.1569, 3.4060, 160 # -0.7565, 0.0247, 0.8477, 0.3714, -0.1043, 1.5607, 161 # 4.0700, -1.8363, 0.4370, -0.3571, 0.7268, 0.3435, 162 # 0.0972, 7.1477, -0.1486, 0.3342, -0.9733, 0.2311, 163 # 0.6104, -0.4988, 2.8838, -1.3387, 1.4291, -0.4121, 164 # -0.6722, 2.6834, -0.5188, 0.0428, -0.3452, -0.0131, 165 # -0.9004, -3.0346, -0.9254, 0.0150, -0.0386, 1.0639, 166 # -0.2444, -0.4562, 4.1626, -1.9304, 1.0662, 2.0683, 167 # -0.9553, -0.6434, -1.6777, 0.0702, -0.0113, -0.5503, 168 # -0.1873, -0.6916, 1.0729, -0.8234, 0.6421, 0.3022, 169 # -0.6065, 0.1016, 0.7792, 0.2533, -0.2670, -0.1314, 170 # 0.7515, -0.9859, 0.5050, -1.0552, -0.5632, 1.0697]], 171 172 # [[ 0.3612, 2.4727, -4.6103, -1.6459, -1.7761, -1.4302, 173 # 0.2737, 0.3302, 4.0617, 0.7206, -0.0749, -0.8146, 174 # -1.0134, 0.8741, 1.9300, 0.5426, 0.1386, -1.3920, 175 # 0.4602, 1.3387, 0.0068, -0.3648, -7.6665, 0.9011, 176 # -0.3286, -1.5220, -0.2155, 0.7959, 4.0746, 0.9382, 177 # 0.0023, -0.2666, 0.4571, -1.9530, 0.3216, 2.1178, 178 # 0.4043, -0.0309, 2.5116, -1.2250, -0.9842, 5.0822, 179 # -4.1296, -5.3579, 0.9115, 0.4843, 0.1365, 0.0491, 180 # 0.1446, 0.6523, 0.0765, 0.3761, 0.0310, -0.4825, 181 # 7.1485, -0.4211, -3.7914, -0.2492, -0.3775, -0.4745, 182 # -1.3320, -1.8203, -1.0266, -0.4446, 2.2385, 0.6003, 183 # -0.1759, -1.9601, 2.3865, 1.3325, 4.8762, -0.2398, 184 # 7.5251, -0.4380, -2.3422, 0.6013, 13.8362, -0.4112, 185 # 2.3579, -0.1720, 1.0265, 0.6521, -0.7363, -0.7864, 186 # 0.2986, -0.1298, 0.5078, 0.1386, 1.4856, -0.3133, 187 # 0.9933, 1.5144, -0.0433, 1.0841, 0.3962, -0.0024, 188 # -0.3937, 2.2719, -0.0198, 1.6771, -1.2469, -0.8017, 189 # 0.1607, 0.0244, -0.1429, 0.9912, 0.1635, -1.2396, 190 # -0.1615, 1.0921, 0.8146, -0.3309, 0.8553, 0.4243, 191 # -3.5547, -0.1382, 0.1513, 0.4036, -0.2505, 0.2295, 192 # -0.6219, -0.7644, -0.7568, -0.4494, -0.0775, -0.0178, 193 # -0.2550, 0.2258, -2.7895, -0.3362, -0.2364, -0.9864, 194 # -1.3459, -1.8118, -0.4397, -0.8312, 0.3526, 1.3541, 195 # -0.0467, 1.6161, -0.4478, -0.5202, -0.4164, -0.8265, 196 # 0.1626, -4.2044, 3.2649, 0.2940, -0.8260, -0.4956]]], device='cuda:0'))
######## Padding functions used in pytorch. #########
1. torch.nn.utils.rnn.PackedSequence(*args)
Holds the data and list of batch_sizes of a packed sequence.
All RNN modules accept packed sequences as inputs.
2. torch.nn.utils.rnn.pack_padded_sequence(input, lengths, batch_first=False)
Packs a Tensor containing padded sequences of variable length.
Input can be of size T*B** where T is the length of the longest sequence, B is the batch size, and the * is any number of dimensions.
If batch_first is True B*T** inputs are expected. The sequences should be sorted by length in a decreasing order.
3. torch.nn.utils.rnn.pad_packed_sequence(sequence, batch_first=False, padding_values=0.0, total_length=None)
Pads a packed batch of variable length sequences.
It is an inverse operation to pack_padded_sequence().
Batch elements will be ordered decreasingly by their length.
Note: total_length is useful to implement the pack sequence -> rnn -> unpack sequence .
Return: Tuple of tensor containing the padded sequence, and a tensor containing the list of lengths of each sequence in the batch.
4. torch.nn.utils.rnn.pad_sequence(sequence, batch_first=False, padding_value=0)
Pad a list of variable length Tensors with zero.
5. torch.nn.utils.rnn.pack_sequence(sequences)
Packs a list of variable length Tensors.
6. Tutorials on these functions.
(1). https://zhuanlan.zhihu.com/p/34418001
(2). https://zhuanlan.zhihu.com/p/28472545
总结起来就是：在利用 recurrent neural network 处理变长的句子序列时，我们可以配套的使用：
torch.nn.utils.rnn.pack_padded_sequence () 来对一个 mini-batch 中的句子进行 padding；
torch.nn.utils.rnn.pad_packed_sequence () 来避免 padding 对句子表示的影响。
===
Stay Hungry，Stay Foolish ...
