Commit Graph
18 Commits
Author SHA1 Message Date
Sam Shleifer b54ef78d0c Bart-CNN (#3059)
`generate` code that produces 99% identical summarizations to fairseq on CNN test data, with caching.
2020-03-02 10:35:53 -05:00
Sam Shleifer 6a37588041 spelling: strictly (#3042) 2020-02-27 10:22:35 -05:00
Sam Shleifer 9df74b8bc4 Delete all mentions of Model2Model (#3019) 2020-02-26 11:36:27 -05:00
Sam Shleifer 129f0604ac Delete untested, broken Model2LSTM (#2968) 2020-02-23 11:28:48 -05:00
Sam Shleifer 92487a1dc0 Bart: fix layerdrop and cached decoder_input_ids for generation (#2969) 2020-02-22 16:25:04 -05:00
Sam Shleifer b5b3445c4f Only use F.gelu for torch >=1.4.0 (#2955)
* Only use F.gelu for torch >=1.4.0

* Use F.gelu for newer torch
2020-02-21 16:10:21 -05:00
Sam Shleifer 53ce3854a1 New BartModel (#2745)
* Results same as fairseq
* Wrote a ton of tests
* Struggled with api signatures
* added some docs
2020-02-20 18:11:13 -05:00
Sam Shleifer 20fc18fbda Skip flaky test_tf_question_answering (#2845)
* Skip flaky test

* Style
2020-02-18 16:14:50 -05:00
Sam Shleifer 0ed630f139 Attempt to increase timeout for circleci slow tests (#2844) 2020-02-13 09:11:03 -05:00
Sam Shleifer ef74b0f07a get_activation('relu') provides a simple mapping from strings i… (#2807)
* activations.py contains a mapping from string to activation function
* resolves some `gelu` vs `gelu_new` ambiguity
2020-02-13 08:28:33 -05:00
sshleifer 5f2dd71d1b Smaller diff 2020-02-11 17:20:09 -05:00
sshleifer 31158af57c formatting 2020-02-11 17:20:09 -05:00
sshleifer 5dd61fb9a9 Add more specific testing advice to Contributing.md 2020-02-11 17:20:09 -05:00
sshleifer 9e5b549b4d fix default getattr 2020-02-04 16:38:52 -05:00
sshleifer 25848a6094 double quotes 2020-02-04 16:38:52 -05:00
sshleifer cbcb83f21d minor cleanup of test_attention_outputs 2020-02-04 16:38:52 -05:00
sshleifer 119610b5c5 Merge branch 'master' into delete-n-special-doc 2019-09-19 01:35:01 -07:00
sshleifer 08e4ad5eea Remove documentation for unused kwarg 2019-09-18 16:35:01 -07:00