add debugging stuff

main.py

@@ -128,7 +128,7 @@ class Model:
  def __init__(self, train_file, test_file, batch_size=32, embedding_size=20, max_norm=40, lr=0.01, num_hops=3, adj_weight_tying=True, **kwargs):
  train_lines, test_lines = self.get_lines(train_file), self.get_lines(test_file)
  lines = np.concatenate([train_lines, test_lines], axis=0)
- vocab, word_to_idx, max_seqlen, max_sentlen = self.get_vocab(lines)
+ vocab, word_to_idx, idx_to_word, max_seqlen, max_sentlen = self.get_vocab(lines)
 
  self.data = { 'train': {}, 'test': {} }
  S_train, self.data['train']['C'], self.data['train']['Q'], self.data['train']['Y'] = self.process_dataset(train_lines, word_to_idx, max_sentlen, offset=0)
@@ -137,7 +137,6 @@ def __init__(self, train_file, test_file, batch_size=32, embedding_size=20, max_
  for i in range(10):
  for k in ['C', 'Q', 'Y']:
  print k, self.data['test'][k][i]
-
  print 'batch_size:', batch_size, 'max_seqlen:', max_seqlen, 'max_sentlen:', max_sentlen
  print 'sentences:', S.shape
  print 'vocab:', len(vocab), vocab
@@ -159,7 +158,8 @@ def __init__(self, train_file, test_file, batch_size=32, embedding_size=20, max_
  self.lb = lb
  self.init_lr = lr
  self.lr = self.init_lr
- self.questions = np.array([x for x in lines if x['type']=='q'])
+ self.S = S
+ self.idx_to_word = idx_to_word
 
  c = T.imatrix()
  q = T.ivector()
@@ -238,7 +238,11 @@ def compute_f1(self, dataset):
  y_true = [self.vocab.index(y) for y in dataset['Y'][:len(y_pred)]]
  print metrics.confusion_matrix(y_true, y_pred)
  print metrics.classification_report(y_true, y_pred)
- return metrics.f1_score(y_true, y_pred, average='weighted', pos_label=None)
+ errors = []
+ for i,(t,p) in enumerate(zip(y_true,y_pred)):
+ if t != p:
+ errors.append((i,self.lb.classes_[p]))
+ return metrics.f1_score(y_true, y_pred, average='weighted', pos_label=None), errors
 
  def train(self, n_epochs=100, shuffle_batch=False):
  epoch = 0
@@ -267,13 +271,27 @@ def train(self, n_epochs=100, shuffle_batch=False):
  print 'epoch:', epoch, 'cost:', (total_cost / len(indices)), ' took: %d(s)' % (end_time - start_time)
 
  print 'TRAIN', '=' * 40
- train_f1 = self.compute_f1(self.data['train'])
+ train_f1, train_errors = self.compute_f1(self.data['train'])
+ for i,pred in train_errors[:10]:
+ print 'context: ', self.to_words(self.data['train']['C'][i])
+ print 'question: ', self.to_words([self.data['train']['Q'][i]])
+ print 'correct answer: ', self.data['train']['Y'][i]
+ print 'predicted answer: ', pred
+ print '---' * 20
+
  print 'TRAIN ERROR:', 1-train_f1
 
  print 'TEST', '=' * 40
- test_f1 = self.compute_f1(self.data['test'])
+ test_f1, test_errors = self.compute_f1(self.data['test'])
  print '*** TEST ERROR:', 1-test_f1
 
+ def to_words(self, indices):
+ sents = []
+ for idx in indices:
+ words = ' '.join([self.idx_to_word[idx] for idx in self.S[idx] if idx > 0])
+ sents.append(words)
+ return ' '.join(sents)
+
  def shuffle_sync(self, dataset):
  p = np.random.permutation(len(dataset['Y']))
  for k in ['C', 'Q', 'Y']:
@@ -310,14 +328,18 @@ def get_vocab(self, lines):
  for w in vocab:
  word_to_idx[w] = len(word_to_idx) + 1
 
+ idx_to_word = {}
+ for w,idx in word_to_idx.iteritems():
+ idx_to_word[idx] = w
+
  max_seqlen = 0
  for i,line in enumerate(lines):
  if line['type'] == 'q':
  id = line['id']-1
  indices = [idx for idx in range(i-id, i) if lines[idx]['type'] == 's']
  max_seqlen = max(len(indices), max_seqlen)
 
- return vocab, word_to_idx, max_seqlen, max_sentlen
+ return vocab, word_to_idx, idx_to_word, max_seqlen, max_sentlen
 
  def process_dataset(self, lines, word_to_idx, max_sentlen, offset):
  S, C, Q, Y = [], [], [], []