for Python3 compatible

fluid/text_matching_on_quora/configs/__init__.py

 
-from cdssm import cdssm_base
-from dec_att import decatt_glove
-from sse import sse_base
-from infer_sent import infer_sent_v1
-from infer_sent import infer_sent_v2
+from .cdssm import cdssm_base
+from .dec_att import decatt_glove
+from .sse import sse_base
+from .infer_sent import infer_sent_v1
+from .infer_sent import infer_sent_v2

fluid/text_matching_on_quora/configs/cdssm.py

 
-import basic_config
+from . import basic_config
 
 def cdssm_base():
     """

fluid/text_matching_on_quora/configs/dec_att.py

 
-import basic_config
+from . import basic_config
 
 def decatt_glove():
     """

fluid/text_matching_on_quora/configs/infer_sent.py

 
-import basic_config
+from . import basic_config
 
 def infer_sent_v1():
     """

fluid/text_matching_on_quora/configs/sse.py

 
-import basic_config
+from . import basic_config
 
 def sse_base():
     """

fluid/text_matching_on_quora/models/__init__.py

-from cdssm import cdssmNet
-from dec_att import DecAttNet
-from sse import SSENet
-from infer_sent import InferSentNet
+from .cdssm import cdssmNet
+from .dec_att import DecAttNet
+from .sse import SSENet
+from .infer_sent import InferSentNet

fluid/text_matching_on_quora/models/infer_sent.py

+
 import paddle.fluid as fluid
-from my_layers import bi_lstm_layer
-from match_layers import ElementwiseMatching
+from .my_layers import bi_lstm_layer
+from .match_layers import ElementwiseMatching
 
 class InferSentNet():
     """

fluid/text_matching_on_quora/models/match_layers.py

@@ -3,7 +3,6 @@ This Module provide different kinds of Match layers
 """
 
 import paddle.fluid as fluid
-import paddle.v2 as paddle
 
 
 def MultiPerspectiveMatching(vec1, vec2, perspective_num):
@@ -44,18 +43,3 @@ def ElementwiseMatching(vec1, vec2):
     return fluid.layers.concat(input=[vec1, vec2, elementwise_mul, elementwise_abs_sub], axis=1)
 
 
-def MultiPerspectiveFullMatching(seq1, seq2, perspective_num):
-    """
-    seq1: Lod tensor with shape [-1, feature_dim] (lod level == 1) is a representation of a sentence.
-    seq2: Another Lod tensor with shape [-1, feature_dim] (lod level == 1) is a representation of a sentence.
-    use seq1 to match seq2
-    return match seq with same shape as seq1.
-    """
-    print seq2
-    seq2_last = fluid.layers.sequence_pool(input=seq2, pool_type="last")
-    print seq2_last
-    seq2 = fluid.layers.sequence_expand(seq2_last, seq1)
-    print seq2
-    #seq2 = fluid.layers.lod_reset(x=seq2, y=seq1)
-    seq2.set_lod(seq1)
-    print seq2

fluid/text_matching_on_quora/models/sse.py

+
 import paddle.fluid as fluid
-from my_layers import bi_lstm_layer
-from match_layers import ElementwiseMatching
+from .my_layers import bi_lstm_layer
+from .match_layers import ElementwiseMatching
 
 class SSENet():
     """

fluid/text_matching_on_quora/pretrained_word2vec.py

@@ -2,27 +2,29 @@
 This Module provide pretrained word-embeddings 
 """
 
-from __future__ import print_function 
+from __future__ import print_function, unicode_literals 
 import numpy as np
 import time, datetime
+import os, sys
+
 
 def Glove840B_300D(filepath, keys=None):
     """
     input: the "glove.840B.300d.txt" file path
     return: a dict, key: word (unicode), value: a numpy array with shape [300]
     """
-    if keys is not None: 
+    if keys is not None:
         assert(isinstance(keys, set))
     print("loading word2vec from ", filepath)
     print("please wait for a minute.")
     start = time.time()
     word2vec = {}
-
     with open(filepath, "r") as f:
         for line in f:
-            info = line.strip().split()
-            # TODO: test python3
-            word = info[0].decode('utf-8')
+            if sys.version_info <= (3, 0): # for python2
+                line = line.decode('utf-8')
+            info = line.strip("\n").split(" ")
+            word = info[0]
             if (keys is not None) and (word not in keys):
                 continue
             vector = info[1:]
@@ -32,6 +34,9 @@ def Glove840B_300D(filepath, keys=None):
     end = time.time()
     print("Spent ", str(datetime.timedelta(seconds=end-start)), " on loading word2vec.")
     return word2vec
- 
+
 if __name__ == '__main__':
-    embed_dict = Glove840B_300D("data/glove.840B.300d.txt")
+    from os.path import expanduser
+    home = expanduser("~")
+    embed_dict = Glove840B_300D(os.path.join(home, "./.cache/paddle/dataset/glove.840B.300d.txt"))
+    exit(0)

fluid/text_matching_on_quora/quora_question_pairs.py

@@ -20,7 +20,7 @@ import tarfile
 import re
 import string
 import random
-import os
+import os, sys
 import nltk
 from os.path import expanduser
 
@@ -43,7 +43,8 @@ COLUMN_COUNT = 4
 
 
 def tokenize(s):
-    s = s.decode('utf-8')
+    if sys.version_info <= (3, 0): # for python2
+        s = s.decode('utf-8')
     if TOKENIZE_METHOD == "nltk":
         return nltk.tokenize.word_tokenize(s)
     elif TOKENIZE_METHOD == "punctuation":
@@ -116,7 +117,7 @@ def build_dict(file_name, cutoff):
 
     dictionary = sorted(word_freq, key=lambda x: (-x[1], x[0]))
     words, _ = list(zip(*dictionary))
-    word_idx = dict(zip(words, xrange(len(words))))
+    word_idx = dict(zip(words, range(len(words))))
     word_idx['<unk>'] = len(words)
     word_idx['<pad>'] = len(words) + 1
     return word_idx

fluid/text_matching_on_quora/train_and_evaluate.py

@@ -9,7 +9,6 @@ import contextlib
 import numpy as np
 
 import paddle.fluid as fluid
-import paddle.v2 as paddle
 
 import utils, metric, configs
 import models
@@ -155,7 +154,7 @@ def train_and_evaluate(train_reader,
 
     # start training
     print("[%s] Start Training" % time.asctime(time.localtime(time.time())))
-    for epoch_id in xrange(global_config.epoch_num):
+    for epoch_id in range(global_config.epoch_num):
         data_size, data_count, total_acc, total_cost = 0, 0, 0.0, 0.0
         batch_id = 0
         for data in train_reader():

fluid/text_matching_on_quora/utils.py

@@ -7,7 +7,7 @@ import time
 import numpy as np
 
 import paddle.fluid as fluid
-import paddle.v2 as paddle
+import paddle
 import quora_question_pairs