Are there instructions on how to generate summaries from provided models?

[RevanthRameshkumar]

Hello,
Are there any instructions on how to use the provided bins/joblibs to generate summaries from a file or string?

Thanks!

[tingofurro]

Hello Revanth,

Sorry for the late reply.
Here's how one could use the pre-trained model to generate a summary:

from model_generator import GeneTransformer

generator = GeneTransformer(device="cuda") # Initialize the generator
generator.reload("/path/to/summary_loop_length_61.bin")
document = "This is a long document I want to summarize"

# Have to put in list because the decode function is meant to be used in batches for efficiency.
# You can use a beam size or not (beam_size), and you can use sampling or not (sample), without sampling it does argmax/top_k

summary = generator.decode([document], max_output_length=61, beam_size=1, return_scores=False, sample=False)

I hope this helps and sorry for the very late answer...
I love the CRD3 dataset by the way!

Philippe

[badrinath-reddy]

Hi,

I could not find "summary_loop_length_61.bin" model in releases,

Thanks

[NightMachinery]

@Badrinath Reddy commented on Jun 1, 2021, 11:17 AM GMT+4:30:

Hi,

I could not find "summary_loop_length_61.bin" model in releases,

Thanks

https://github.com/CannyLab/summary_loop/releases/download/v0.1/summary_loop_length46.bin is there though.