pointer-generator

pointer-generator copied to clipboard

Hi, Thanks for great work! Is there any way to run single_pass without installing ROUGE? Thanks.

jonaseltes

non-Coverage Mechanism VS Coverage Mechanism

1

I have changed the word-attention mechanism in pointer generator. I have trained my model for 240k iterations without coverage. Then for 4k iterations i trained with coverage the decoded results...

AliRaza1844

i put the coverage loss in my program,and i get the coverage loss, always 0.89? why?

yanshengli

Provided test set output is different from the output generated using the pre-trained models.

6

1. I used the pre-trained models (TF 1.2.1) to decode and calculated ROUGE scores for those. These outputs look different from the ones provided under the test_output folder. For example:...

LeenaShekhar

Multiple times article and result summary printed in debug mode

1

When i am trying to run in debug mode, an article from test set along with summary is printed multiple times. What is the logic behind the same ?

lakshyakh93

Difference in paper and codes.

1

https://github.com/abisee/pointer-generator/blob/a7317f573d01b944c31a76bde7218bcfc890ef6a/attention_decoder.py#L173 Hi, abisee! I am confused about the codes in this line. In your paper, the function is V[s_t, h*_t] + b But in the codes, you use rnn_output instead....

caoxu915683474

Hi, abisee! Thanks for sharing this good work! I have some question about your paper in Experiments part. You said that with the coverage loss ratio set to 1.0 the...

caoxu915683474

TF Flag, for mode is raising exception --mode="train"

4

```python python run_summarization.py --mode="train" --data_path="../misc/finished_files/chunked/train_*" --vocab_path="../misc/finished_files/vocab" --log_root="log/" --exp_name="first_test" /home/ubuntu/anaconda3/lib/python3.6/site-packages/h5py/__init__.py:36: FutureWarning: Conversion of the second argument of issubdtype from `float` to `np.floating` is deprecated. In future, it will be treated as...

Adityagrao

The loss decreases too fast using LCSTS dataset

7

Hi~ Using the CNN/DailyMail dataset this system works well. Then I just change dataset, create .bin files with make_datafiles.py and remain the other parameters, but the training log shows that...

lixiang1991

Cloud I use Transfer learning for new data?

1

I want to use the pre-trained model for Transfer learning Can we do it? If we can,how to do? Thank you!

dondon2475848