5

Reading the Tensorflow text summarization documentation setup : https://github.com/tensorflow/models/tree/master/textsum it states clone the code to your workspace and create empty WORKSPACE file.

I've created a new folder and cloned https://github.com/tensorflow/tensorflow.git to this folder , is that what is meant by 'clone the code to your workspace' ? What is an empty workspace file ? Is it an empty file named 'WORKSPACE' ?

blue-sky
  • 51,962
  • 152
  • 427
  • 752

1 Answers1

10

WORKSPACE is a file that bazel (tensorflow's build system) searches in the directory hierarchy to determine the root of the project.

You can create a simple empty WORKSPACE file.

$touch WORKSPACE

in my setup I just created another directory 

cd models
mkdir traintextsum
cd traintextsum
ln -sf ../textsum/ .
mkdir data 
touch WORKSPACE
bazel build -c opt --config=cuda textsum/...

keep in mind that the model is not trained. So to produce any meaningful result you would have to have some dataset that has the Gigaword Dataset Format to train it on.

The dataset has a license and thus is not freely available online (costs around $6K or $27K for a year subscription).

I am not aware of any other meaningful dataset that has that format besides the Gigaword itself.

fabrizioM
  • 46,639
  • 15
  • 102
  • 119
  • 2
    I can use the model without Gigaword because ? : https://github.com/tensorflow/models/blob/master/textsum/README.md states that 'In How To Run below, users can use toy data and vocab provided in the data/ directory to run the training by replacing the data directory flag.' so replacing the data directory flag gives : bazel-bin/textsum/seq2seq_attention \ --mode=train \ --article_key=article \ --abstract_key=abstract \ --data_path=data/data \ --vocab_path=data/vocab \ --log_root=textsum/log_root \ --train_dir=textsum/log_root/train – blue-sky Aug 30 '16 at 09:53
  • @blue-sky so it seesm there is no way to make it works since it is not pre-trained and the `training set` that Google has used is apparently not free. – loretoparisi Oct 28 '16 at 11:04
  • I get this warning: "WARNING: Config values are not defined in any .rc file: cuda" after running "bazel build". Thoughts? – vgoklani Dec 20 '16 at 16:46
  • @blue-sky "users can use toy data and vocab" how did it go for you did you manage to create a model and run was it a good model? – Jas Jul 06 '18 at 04:24