search

Yale-LILY / SummerTime

An open-source text summarization toolkit for non-experts. EMNLP'2021 Demo
https://arxiv.org/abs/2108.12738
Apache License 2.0
268 stars 30 forks source link

Yusen hmnet1 #33

Closed yszh8 closed 3 years ago

yszh8 commented 3 years ago

This is an intermediate result for HMNet. We need to merge after pipelining the QMSum dataset etc.

TODOs:

  1. checkpoint saving and loading
  2. pos_tag and role vector saving
  3. interface to the "corpus"
  4. minimize the dependencies that need to be installed
niansong1996 commented 3 years ago

@chatc Does the new commit fix the spacy version issues?

yszh8 commented 3 years ago

@chatc Does the new commit fix the spacy version issues?

Yes, now we can run HMNet using SpaCy 3! (although HMNet has not been pushed to our repo yet)

niansong1996 commented 3 years ago

@chatc Where are we on this thread? Have you checked how to make a repo as submodules?

niansong1996 commented 3 years ago

@chatc Nice! But are we sure that the submodule is correct and working? Asking because I checked https://github.com/Yale-LILY/SummerTime/tree/yusen_hmnet1/model/third_party and clicked the link and it gives me 404

yszh8 commented 3 years ago

@chatc Nice! But are we sure that the submodule is correct and working? Asking because I checked https://github.com/Yale-LILY/SummerTime/tree/yusen_hmnet1/model/third_party and clicked the link and it gives me 404

I found this issue as well. I am not sure if this is because I modified the submodule code after adding it, but I can't modify the submodule's remote repo, i.e. HMNet.

niansong1996 commented 3 years ago

We've decided to give green light for the code from HMNet to be copied for you to import it easily.

Meanwhile, can you make a PR or Issue on the HMNet page and let them know that their spacy version is outdated?

niansong1996 commented 3 years ago

I don't think I can review the 110 files changed... So I can only trust you on this ;)

Do you have tests and is the code passing them?

yszh8 commented 3 years ago

I don't think I can review the 110 files changed... So I can only trust you on this ;)

Do you have tests and is the code passing them?

Yeah, I tested the model on our dataset, and passed it! I can test more cases after Troy's integration test.