Commits · 629aac92ec33c664b8470221db0a3d1921ccdc55 · chenpangpang / transformers

11 Mar, 2020 11 commits
- finalized PR · d880a5fb
  patrickvonplaten authored Mar 07, 2020
  
  d880a5fb
- best current version and make style · 2acfe639
  patrickvonplaten authored Mar 06, 2020
  
  2acfe639
- fix conflicts · c62444da
  patrickvonplaten authored Mar 08, 2020
  
  c62444da
- add current changes · 77e67750
  Patrick von Platen authored Mar 06, 2020
  
  77e67750
- comment out stuff · 42121699
  Patrick von Platen authored Mar 06, 2020
  
  42121699
- work in progress · 7a11e925
  Patrick von Platen authored Mar 06, 2020
  
  7a11e925
- only do output_past=True for language generation in bart · aceb3fba
  Patrick von Platen authored Mar 05, 2020
  
  aceb3fba
- better naming · 7cba11fb
  Patrick von Platen authored Mar 05, 2020
  
  7cba11fb
- fix conflicts · ff648221
  Patrick von Platen authored Mar 06, 2020
  
  ff648221
- refactored code a bit and made more generic · c0d9dd3b
  Patrick von Platen authored Mar 05, 2020
  
  c0d9dd3b
- fix conflicts · d8e2b3c5
  Patrick von Platen authored Mar 06, 2020
  
  d8e2b3c5
06 Mar, 2020 1 commit
- [Bart] _prepare_decoder_inputs should use large negative (#3158) · ed37f9fa
  Sam Shleifer authored Mar 06, 2020
  
  ed37f9fa
05 Mar, 2020 4 commits
- fix renaming problem · 58fc8f97
  patrickvonplaten authored Mar 06, 2020
  
  58fc8f97
- Rename BartForMaskedLM -> BartForConditionalGeneration (#3114) · 857e0a0d
  Sam Shleifer authored Mar 05, 2020
```
* improved documentation
```
  857e0a0d
- cleanup deltas · 1360daca
  sshleifer authored Mar 05, 2020
  
  1360daca
- tests pass · c36fdc88
  sshleifer authored Mar 05, 2020
  
  c36fdc88
03 Mar, 2020 1 commit
- BartForSequenceClassification: fix num_labels, add test (#3110) · e9e6efdc
  Sam Shleifer authored Mar 03, 2020
  
  e9e6efdc
02 Mar, 2020 1 commit

Sam Shleifer authored Mar 02, 2020

`generate` code that produces 99% identical summarizations to fairseq on CNN test data, with caching.

b54ef78d

26 Feb, 2020 2 commits
- Fix bart slow test · f5516805
  Julien Chaumond authored Feb 26, 2020
  
  f5516805
- Fix (non-slow) tests on GPU (torch) (#3024) · 9cda3620
  Julien Chaumond authored Feb 26, 2020
```
* Fix tests on GPU (torch)

* Fix bart slow tests
Co-authored-by: Sam Shleifer <sshleifer@gmail.com>
```
  9cda3620
22 Feb, 2020 1 commit
- Bart: fix layerdrop and cached decoder_input_ids for generation (#2969) · 92487a1d
  Sam Shleifer authored Feb 22, 2020
  
  92487a1d
20 Feb, 2020 1 commit

New BartModel (#2745) · 53ce3854

Sam Shleifer authored Feb 20, 2020

* Results same as fairseq
* Wrote a ton of tests
* Struggled with api signatures
* added some docs

53ce3854