Procedure to generate text using the pretrained model

[TinoRed]

Hi, I wanted to try the pretrained model aug17-80.pth on a local CSV file to see if it suits my needs without going through a new training.

I followed this procedure :

• Refactored the title file modifying utils/refactorTitles.py script specifying the path to my txt title file ;
• Modified the etc/templatePreprocess.py script, to obtain the necessary files to launch model/summarize.py ( in particular testData.txt and testTitle.txt
• Launched model/summarize.py specifiying the correct paths to my files.
• Modified etc/summaryComparison.py , to substitute values in template.

Using this procedure produces some output but it is still necessary to provide a captions file.

My question is : Could you provide a procedure to test the aug17-80.pth model on a local CSV file, only providing the csv file and its relative title file, without providing a captions file?

I think it would be very useful for future users to try the pretrained model on their data in an easy and controlled way, without training a model from scratch.

[JasonObeid]

Hey @TinoRed, this sounds like a good idea, but I unfortunately don't have the time to work on this.

If it's something you would want to work on, feel free to submit a pull request. I could try to answer any of your potential questions :)