How to use the pretrain model?

[zhhengcs]

if i want to use the model pretrained,which parameter shold i set and how?

[mohman22]

1. Download the CNN data
2. Download the pointer-generator-master
3. Download the pretrained model and put into same directory as pointer-generator-master
4. Run the line below (chunked data is from CNN, vocab is also from CNN directory, log_root is any directory you make to save logs, experiment name is "myexperiment"

python run_summarization.py --mode=train --data_path=/path/to/chunked/train_* --vocab_path=/path/to/vocab --log_root=/path/to/a/log/directory --exp_name=myexperiment

[scigeek72]

@mohman22 don't you just need to run the decoder portion (--mode=decode ) of the code if you are using only the pre-trained model?

[yomnaa]

download the pretrained zip file extract it and open the readme file you will find this command at the end of the readme file

If you'd like to see randomly-generated examples from the validation set print to screen, run:

python run_summarization.py --mode=decode --data_path=/path/to/data/val_* --vocab_path=/path/to/data/vocab --log_root=/path/to/directory/containing/pretrained_model --exp_name=pretrained_model --max_enc_steps=400 --max_dec_steps=120 --coverage=1

after extracting the zip file ...there will be folder with the same name as the zip file..rename it to "pretrained_model" and in the command replace the log_root path with the path of the pretrained_model folder, for exmple if the path of the renamed pretrained_model folder was ---> "/ss/dd/pretrained_model" ...... the command would be --log_root="/ss/dd". The point is that code append the exp_name to the log_root and make it the path to look for a folder named train and looks for checkpoint inside the train folder.

[Ann1583]

I have followed the steps mentioned here to run the pretrained model but i am getting this error "TypeError: unsupported operand type(s) for *: 'int' and 'Flag'" What should I do?

[yomnaa]

Which tensor fow version are you using?

[Ann1583]

I checked the version it was 1.7.0 . I have installed 1.2.1 and now i am getting this error AssertionError: Error: Empty filelist at ../finished_files/val_*

I have checked this file but it has content in it

[yomnaa]

Have you prepared the dataset the way written in readme file of the repo?

On Tue, Apr 10, 2018, 6:53 AM Ann1583 [email protected] wrote:

I checked the version it was 1.7.0 . I have installed 1.2.1 and now i am getting this error AssertionError: Error: Empty filelist at ../finished_files/val_*

I have checked this file but it has content in it

— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/abisee/pointer-generator/issues/77#issuecomment-379973345, or mute the thread https://github.com/notifications/unsubscribe-auth/ARqRdiZ2mVgTbpTJtSg08uZmrMaUPgHkks5tnDrEgaJpZM4R4H0T .

[RaiAbhishek]

Refer your file "run_summarization.py" and this: https://stackoverflow.com/questions/49206318/typeerror-unsupported-operand-types-for-int-and-flag/49734816#49734816 Works fine with newer tf versions.

[Ann1583]

@yomnaa I actually downloaded the processed data files provided by JafferWilson https://github.com/JafferWilson/Process-Data-of-CNN-DailyMail

Should I process the files myself? I was having trouble with Stanford CoreNLP, that's why I switched to these processed files.

[Namrata96]

I have followed the steps mentioned here to run the pretrained model but i am getting this error "TypeError: unsupported operand type(s) for *: 'int' and 'Flag'" What should I do?

@Ann1583 If you still haven't solved this : Just replace https://github.com/abisee/pointer-generator/blob/879cdb3e0dd1d17210d0e24f545c843c33a39cef/run_summarization.py#L299 with hps_dict[key] = val.value # add it to the dict. This works because val is a flag object, and val.value gives you the value of the flag. Hope this helps!

[yasersakkaf]

In the command what does the following parameter refer to? --data_path=/path/to/data/val_* Please can anyone explain. Thanks in Advance

[shivam13juna]

@yasersakkaf path to validation data?

[jawalemohit]

I have followed the steps mentioned here to run the pretrained model but i am getting this error "TypeError: unsupported operand type(s) for *: 'int' and 'Flag'" What should I do?

use .value to your batch.py file on the line it is giving error

[shivam13juna]

or you can use my updated repo to tensorflow 1.12, that oughta take care of it all, here

[ojassm]

@yasersakkaf I guess it refers to the validation dataset i.e the val.bin file in the finished files if you're using the given dataset

[husongjiang]

I have followed the steps mentioned here to run the pretrained model but i am getting this error "TypeError: unsupported operand type(s) for *: 'int' and 'Flag'" What should I do?

.vlaue 能够将flag转换为int