mast-group / convolutional-attention Goto Github PK

OS: Ubuntu
command: python copy_conv_rec_learner.py model 10 128 test_result

  File "copy_conv_rec_learner.py", line 11, in <module>
      from experimenter import ExperimentLogger
  ImportError: No module named experimenter

WARNING (theano.tensor.blas): Using NumPy C-API based implementation for BLAS functions.
Extracting data...
/home/sawan/convnet/convolutional-attention-master/convolutional_attention/conv_att_rec_model.py:133: UserWarning: The image_shape keyword argument to tensor.nnet.conv2d is deprecated, it has been renamed to input_shape.
filter_shape=self.conv_layer1_code.get_value().shape)
/home/sawan/convnet/convolutional-attention-master/convolutional_attention/conv_att_rec_model.py:138: UserWarning: The image_shape keyword argument to tensor.nnet.conv2d is deprecated, it has been renamed to input_shape.
filter_shape=self.conv_layer2_code.get_value().shape)
/home/sawan/convnet/convolutional-attention-master/convolutional_attention/conv_att_rec_model.py:152: UserWarning: The image_shape keyword argument to tensor.nnet.conv2d is deprecated, it has been renamed to input_shape.
filter_shape=self.conv_layer3_code.get_value().shape)[:, 0, :, 0]
Traceback (most recent call last):
File "convolutional_attention/conv_att_rec_learner.py", line 235, in
model.train(input_file, max_epochs=max_num_epochs)
File "convolutional_attention/conv_att_rec_learner.py", line 35, in train
self.naming_data.name_empirical_dist)
File "/home/sawan/convnet/convolutional-attention-master/convolutional_attention/conv_att_rec_model.py", line 18, in init
self.__init_parameter(empirical_name_dist)
File "/home/sawan/convnet/convolutional-attention-master/convolutional_attention/conv_att_rec_model.py", line 88, in __init_parameter
self.__compile_model_functions()
File "/home/sawan/convnet/convolutional-attention-master/convolutional_attention/conv_att_rec_model.py", line 227, in __compile_model_functions
self.grad_step = theano.function(inputs=[], updates=step_updates, outputs=ratios)
File "/usr/local/lib/python2.7/dist-packages/theano/compile/function.py", line 317, in function
output_keys=output_keys)
File "/usr/local/lib/python2.7/dist-packages/theano/compile/pfunc.py", line 449, in pfunc
no_default_updates=no_default_updates)
File "/usr/local/lib/python2.7/dist-packages/theano/compile/pfunc.py", line 208, in rebuild_collect_shared
raise TypeError(err_msg, err_sug)
TypeError: ('An update must have the same type as the original shared variable (shared_var=sentence_count, shared_var.type=TensorType(float64, scalar), update_val=Alloc.0, update_val.type=TensorType(float64, (True,))).', 'If the difference is related to the broadcast pattern, you can call the tensor.unbroadcast(var, axis_to_unbroadcast[, ...]) function to remove broadcastable dimensions.')

Hello!

I used the model in copy_conv_rec_learner.py in order to train on the method names of round about 160000 methods. The training for 30 epochs took round about a day. However, the calculation of the scores after the model was saved wasn't finished after 8 hours. The validation dataset covered less than 20000 methods.

The json-files were structured acording to the description here:

[
  {"tokens":["<SENTENCE_START>","{","return","project","params",";","}","<SENTENCE_END/>"], "name":["get","project","params"]},
  ...
]

Therefore my question: is the input json file wrong or is it normal for the scoring to take that long?

Thanks in advance!

Hi there,

I really like your work and I'm trying to work out something based on your project.
I saw your paper mention that you used another baseline model, which is a standard attention model (Bahdanau et al., 2015) with biRNN and fully connected components. But I didn't see the source code in your project here.

Would you please help me with it?

Thanks,
Rui

TypeError Traceback (most recent call last)
D:\Project\convolutional-attention-master\convolutional_attention\baseline.py in ()
85 res = [token_precision_recall(name, real_name) for name in suggested_names]
86 is_correct = [len(real_name) == len(name) and all(k == l for k, l in zip(name, real_name)) for name in suggested_names]
---> 87 eval.add_result([1.] * num_ranks , is_correct, [False] * num_ranks, res)
88 print ''
89 print eval

TypeError: add_result() takes exactly 6 arguments (5 given)