Skip to content
This repository has been archived by the owner on Jan 15, 2024. It is now read-only.

[MODEL] XLNet conversion scripts #866

Merged
merged 19 commits into from
Oct 16, 2019
Merged

[MODEL] XLNet conversion scripts #866

merged 19 commits into from
Oct 16, 2019

Conversation

leezu
Copy link
Contributor

@leezu leezu commented Aug 7, 2019

Description

Add XLNet conversion scripts

Checklist

Essentials

  • PR's title starts with a category (e.g. [BUGFIX], [MODEL], [TUTORIAL], [FEATURE], [DOC], etc)
  • Changes are complete (i.e. I finished coding on this PR)
  • All changes have test coverage
  • Code is well-documented

Changes

  • Pretrained XLNet

Comments

This PR resolves #787

@leezu leezu requested a review from szha as a code owner August 7, 2019 17:34
@codecov
Copy link

codecov bot commented Aug 7, 2019

Codecov Report

Merging #866 into master will increase coverage by 15.86%.
The diff coverage is n/a.

Impacted file tree graph

@@             Coverage Diff             @@
##           master     #866       +/-   ##
===========================================
+ Coverage   73.87%   89.74%   +15.86%     
===========================================
  Files          67       67               
  Lines        6423     6423               
===========================================
+ Hits         4745     5764     +1019     
+ Misses       1678      659     -1019
Impacted Files Coverage Δ
src/gluonnlp/data/utils.py 74.04% <ø> (ø) ⬆️
src/gluonnlp/data/batchify/embedding.py 47.69% <0%> (-50.01%) ⬇️
src/gluonnlp/vocab/subwords.py 84.78% <0%> (-2.18%) ⬇️
src/gluonnlp/data/batchify/batchify.py 96.06% <0%> (+0.78%) ⬆️
src/gluonnlp/data/transforms.py 86% <0%> (+1.11%) ⬆️
src/gluonnlp/data/question_answering.py 100% <0%> (+1.66%) ⬆️
src/gluonnlp/model/train/embedding.py 87.17% <0%> (+2.56%) ⬆️
src/gluonnlp/vocab/elmo.py 96.66% <0%> (+3.33%) ⬆️
src/gluonnlp/data/corpora/wikitext.py 100% <0%> (+5.17%) ⬆️
src/gluonnlp/embedding/token_embedding.py 91.96% <0%> (+5.69%) ⬆️
... and 21 more

@mli
Copy link
Member

mli commented Aug 7, 2019

Job PR-866/1 is complete.
Docs are uploaded to http://gluon-nlp-staging.s3-accelerate.dualstack.amazonaws.com/PR-866/1/index.html

@mli
Copy link
Member

mli commented Aug 8, 2019

Job PR-866/2 is complete.
Docs are uploaded to http://gluon-nlp-staging.s3-accelerate.dualstack.amazonaws.com/PR-866/2/index.html

@mli
Copy link
Member

mli commented Aug 16, 2019

Job PR-866/4 is complete.
Docs are uploaded to http://gluon-nlp-staging.s3-accelerate.dualstack.amazonaws.com/PR-866/4/index.html

@mli
Copy link
Member

mli commented Aug 16, 2019

Job PR-866/5 is complete.
Docs are uploaded to http://gluon-nlp-staging.s3-accelerate.dualstack.amazonaws.com/PR-866/5/index.html

@eric-haibin-lin
Copy link
Member

Is this ready?

@leezu
Copy link
Contributor Author

leezu commented Aug 17, 2019

This only contains the model and no fine-tuning scripts. For finetuning, designing a joint API encompasses the existing bert scripts and works for XLNet and moving that to the main package would be a way forward. That can be done in a separate PR though I guess.

Copy link
Member

@eric-haibin-lin eric-haibin-lin left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I don't see an example usage of XLNet with tokenizers and how I can get it with xx.get_model API. Are you going to add it?

@szha
Copy link
Member

szha commented Sep 2, 2019

@leezu pinging for an update

@szha
Copy link
Member

szha commented Sep 11, 2019

@leezu @eric-haibin-lin gentle ping

@leezu leezu requested a review from a team as a code owner September 25, 2019 21:34
@leezu
Copy link
Contributor Author

leezu commented Oct 8, 2019

@mli
Copy link
Member

mli commented Oct 8, 2019

Job PR-866/7 is complete.
Docs are uploaded to http://gluon-nlp-staging.s3-accelerate.dualstack.amazonaws.com/PR-866/7/index.html

@mli
Copy link
Member

mli commented Oct 8, 2019

Job PR-866/8 is complete.
Docs are uploaded to http://gluon-nlp-staging.s3-accelerate.dualstack.amazonaws.com/PR-866/8/index.html

@mli
Copy link
Member

mli commented Oct 8, 2019

Job PR-866/9 is complete.
Docs are uploaded to http://gluon-nlp-staging.s3-accelerate.dualstack.amazonaws.com/PR-866/9/index.html

@mli
Copy link
Member

mli commented Oct 8, 2019

Job PR-866/10 is complete.
Docs are uploaded to http://gluon-nlp-staging.s3-accelerate.dualstack.amazonaws.com/PR-866/10/index.html

@mli
Copy link
Member

mli commented Oct 8, 2019

Job PR-866/11 is complete.
Docs are uploaded to http://gluon-nlp-staging.s3-accelerate.dualstack.amazonaws.com/PR-866/11/index.html

@szha szha added the release focus Progress focus for release label Oct 8, 2019
Currently unused, but included for future-compatibility of parameter files
We can't create a sentencepiece model from the vocabulary alone. As long as
GluonNLP does not reimplement sentencepiece tokenization, the binary model needs
to distributed as well.
@mli
Copy link
Member

mli commented Oct 9, 2019

Job PR-866/12 is complete.
Docs are uploaded to http://gluon-nlp-staging.s3-accelerate.dualstack.amazonaws.com/PR-866/12/index.html

@mli
Copy link
Member

mli commented Oct 10, 2019

Job PR-866/13 is complete.
Docs are uploaded to http://gluon-nlp-staging.s3-accelerate.dualstack.amazonaws.com/PR-866/13/index.html

@mli
Copy link
Member

mli commented Oct 10, 2019

Job PR-866/15 is complete.
Docs are uploaded to http://gluon-nlp-staging.s3-accelerate.dualstack.amazonaws.com/PR-866/15/index.html

@leezu
Copy link
Contributor Author

leezu commented Oct 10, 2019

CI passes now

@leezu leezu merged commit 2b18e51 into dmlc:master Oct 16, 2019
@leezu leezu deleted the xlnet branch October 16, 2019 22:48
Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.
Labels
release focus Progress focus for release
Projects
None yet
Development

Successfully merging this pull request may close these issues.

Port XLNet to gluonnlp
4 participants