maluuba / mctest-model Goto Github PK

Title^

I follow the instructions and end up with the following errors:
Using Theano backend.
init dataset with h5 file.
(u'input_answer_shape:', (600, 4, 24))
(u'input_dep_shape:', (600, 51, 4))
(u'input_dep_2gram_shape:', (600, 50, 4))
(u'input_dep_3gram_shape:', (600, 49, 4))
(u'input_negation_questions_shape:', (600, 4))
(u'input_question_shape:', (600, 19))
(u'input_question_answer_shape:', (600, 4, 26))
(u'input_reordered_story_shape:', (600, 51, 83))
(u'input_reordered_story_2gram_shape:', (600, 50, 98))
(u'input_reordered_story_3gram_shape:', (600, 49, 141))
(u'input_reordered_story_attentive_shape:', (600, 632))
(u'input_story_shape:', (600, 51, 83))
(u'input_story_2gram_shape:', (600, 50, 98))
(u'input_story_3gram_shape:', (600, 49, 141))
(u'input_story_attentive_shape:', (600, 632))
(u'y_hat_shape:', (600, 4))
INFO:__main__:finish init dataset with data.h5
INFO:embeddings:load mode=in-memory, embedding data type=<type 'numpy.ndarray'>, shape=(3000000, 300)
loading word embedding.
Traceback (most recent call last):
File "run.py", line 92, in <module>
train_option(EPOCHS=args.epoch)
File "run.py", line 72, in train_option
graph = model.build()
File "/Users/allanjie/Documents/workspace/mctest-model/model.py", line 624, in build
self._add_model()
File "/Users/allanjie/Documents/workspace/mctest-model/model.py", line 417, in _add_model
trainable=config['trainable_sentence_ensemble'])(layers_s_input)
File "/Users/allanjie/Documents/workspace/mctest-model/layers.py", line 590, in __init__
self.weights = weights if len(weights) == layers_len else [1.] * layers_len
AttributeError: can't set attribute

the test data in "data.h5" seems to come from MCtest-500,
Where is the test data of MCtest-160?
Do you have codes that can convert these dataset to h5 format?

python2.7 run.py
Using TensorFlow backend.
init dataset with h5 file.
(u'input_answer_shape:', (600, 4, 24))
(u'input_dep_shape:', (600, 51, 4))
(u'input_dep_2gram_shape:', (600, 50, 4))
(u'input_dep_3gram_shape:', (600, 49, 4))
(u'input_negation_questions_shape:', (600, 4))
(u'input_question_shape:', (600, 19))
(u'input_question_answer_shape:', (600, 4, 26))
(u'input_reordered_story_shape:', (600, 51, 83))
(u'input_reordered_story_2gram_shape:', (600, 50, 98))
(u'input_reordered_story_3gram_shape:', (600, 49, 141))
(u'input_reordered_story_attentive_shape:', (600, 632))
(u'input_story_shape:', (600, 51, 83))
(u'input_story_2gram_shape:', (600, 50, 98))
(u'input_story_3gram_shape:', (600, 49, 141))
(u'input_story_attentive_shape:', (600, 632))
(u'y_hat_shape:', (600, 4))
INFO:main:finish init dataset with data.h5
INFO:embeddings:load mode=in-memory, embedding data type=<type 'numpy.ndarray'>, shape=(3000000, 300)
loading word embedding.
2018-11-20 15:09:13.013095: W tensorflow/core/platform/cpu_feature_guard.cc:45] The TensorFlow library wasn't compiled to use SSE4.1 instructions, but these are available on your machine and could speed up CPU computations.
2018-11-20 15:09:13.013120: W tensorflow/core/platform/cpu_feature_guard.cc:45] The TensorFlow library wasn't compiled to use SSE4.2 instructions, but these are available on your machine and could speed up CPU computations.
2018-11-20 15:09:13.013126: W tensorflow/core/platform/cpu_feature_guard.cc:45] The TensorFlow library wasn't compiled to use AVX instructions, but these are available on your machine and could speed up CPU computations.
2018-11-20 15:09:13.013130: W tensorflow/core/platform/cpu_feature_guard.cc:45] The TensorFlow library wasn't compiled to use AVX2 instructions, but these are available on your machine and could speed up CPU computations.
2018-11-20 15:09:13.013135: W tensorflow/core/platform/cpu_feature_guard.cc:45] The TensorFlow library wasn't compiled to use FMA instructions, but these are available on your machine and could speed up CPU computations.
Traceback (most recent call last):
File "run.py", line 92, in
train_option(EPOCHS=args.epoch)
File "run.py", line 72, in train_option
graph = model.build()
File "/u/scratch1/rpujari/parallel_hierarchical/model.py", line 623, in build
self._add_sentence_encode(encode=self.model_config['layers']['sentence_encode']['type'])
File "/u/scratch1/rpujari/parallel_hierarchical/model.py", line 246, in _add_sentence_encode
nodes['__story_encoding1'] = weighting_layer([nodes['story_word_embedding1'], nodes['__w_story1']])
File "/homes/rpujari/anaconda3/envs/flower/lib/python2.7/site-packages/keras/engine/topology.py", line 485, in call
self.add_inbound_node(inbound_layers, node_indices, tensor_indices)
File "/homes/rpujari/anaconda3/envs/flower/lib/python2.7/site-packages/keras/engine/topology.py", line 543, in add_inbound_node
Node.create_node(self, inbound_layers, node_indices, tensor_indices)
File "/homes/rpujari/anaconda3/envs/flower/lib/python2.7/site-packages/keras/engine/topology.py", line 153, in create_node
output_tensors = to_list(outbound_layer.call(input_tensors, mask=input_masks))
File "/u/scratch1/rpujari/parallel_hierarchical/layers.py", line 282, in call
lay1 = T.addbroadcast(lay1, lay1.ndim - 1)
AttributeError: 'Tensor' object has no attribute 'ndim'

Hi,

Could you please publish the scripts to produce the input feature vectors directly from the MCTest data set?

In particular there are features that are hard to reconcile with the paper (sorry if I missed something big). For example the input_dep ?

Following the instructions, using a python2.7 anaconda distribution with Keras 1.1.1 and Then 0.8.2 I get the following failure.

(python2) user@localhost:maluuba/mctest-model$ time python run.py
Using Theano backend.
init dataset with h5 file.
(u'input_answer_shape:', (600, 4, 24))
(u'input_dep_shape:', (600, 51, 4))
(u'input_dep_2gram_shape:', (600, 50, 4))
(u'input_dep_3gram_shape:', (600, 49, 4))
(u'input_negation_questions_shape:', (600, 4))
(u'input_question_shape:', (600, 19))
(u'input_question_answer_shape:', (600, 4, 26))
(u'input_reordered_story_shape:', (600, 51, 83))
(u'input_reordered_story_2gram_shape:', (600, 50, 98))
(u'input_reordered_story_3gram_shape:', (600, 49, 141))
(u'input_reordered_story_attentive_shape:', (600, 632))
(u'input_story_shape:', (600, 51, 83))
(u'input_story_2gram_shape:', (600, 50, 98))
(u'input_story_3gram_shape:', (600, 49, 141))
(u'input_story_attentive_shape:', (600, 632))
(u'y_hat_shape:', (600, 4))
INFO:__main__:finish init dataset with data.h5
INFO:embeddings:load mode=in-memory, embedding data type=<type 'numpy.ndarray'>, shape=(3000000, 300)
loading word embedding.
Traceback (most recent call last):
  File "run.py", line 92, in <module>
    train_option(EPOCHS=args.epoch)
  File "run.py", line 72, in train_option
    graph = model.build()
  File "mctest-model/model.py", line 624, in build
    self._add_model()
  File "mctest-model/model.py", line 417, in _add_model
    trainable=config['trainable_sentence_ensemble'])(layers_s_input)
  File "mctest-model/layers.py", line 594, in __init__
    self.weights = weights if len(weights) == layers_len else [1.] * layers_len
AttributeError: can't set attribute