Commit Graph
34 Commits
Author SHA1 Message Date
Thomas Wolf 3814e167d9 Merge pull request #3225 from patrickvonplaten/finalize_merge_bart_generate_into_default_generate
Complete merge Seq-2-Seq generation into default generation
2020-03-14 15:08:59 +01:00
Sam Shleifer 2bd79e23de [BART] FP16 testing fixes (#3266) 2020-03-13 19:48:26 -04:00
Patrick von Platen 6a82f774f2 fix typo 2020-03-12 21:10:51 +01:00
Patrick von Platen f1c71da115 fix eos_token_ids in test 2020-03-12 21:00:54 +01:00
Patrick von Platen 6047f46b19 re-add eos token to get good bart results 2020-03-12 20:17:50 +01:00
Patrick von Platen ac303eae46 fix problem with half 2020-03-11 12:24:30 +01:00
Patrick von Platen bc9d5d917c make all tensors half precision 2020-03-11 12:15:38 +01:00
Patrick von Platen a332cc9f7f finalize generation merge 2020-03-11 11:53:36 +01:00
Patrick von Platen 7351a8dbaf re-add scoring filtering 2020-03-11 11:06:56 +01:00
Patrick von Platen 374deef48d fixed typo 2020-03-11 11:06:56 +01:00
patrickvonplaten 41b437ea3a add draft version of propsoed changes for ROGUE score 2020-03-11 11:06:56 +01:00
patrickvonplaten a5751f7578 fix bug with attention_mask as optional input argument 2020-03-11 11:06:56 +01:00
patrickvonplaten d880a5fbde finalized PR 2020-03-11 11:06:56 +01:00
patrickvonplaten 2acfe63964 best current version and make style 2020-03-11 11:06:56 +01:00
patrickvonplaten c62444da39 fix conflicts 2020-03-11 11:06:56 +01:00
Patrick von Platen 77e6775065 add current changes 2020-03-11 11:06:56 +01:00
Patrick von Platen 421216997b comment out stuff 2020-03-11 11:06:56 +01:00
Patrick von Platen 7a11e925cf work in progress 2020-03-11 11:06:56 +01:00
Patrick von Platen aceb3fbaf4 only do output_past=True for language generation in bart 2020-03-11 11:06:56 +01:00
Patrick von Platen 7cba11fb9b better naming 2020-03-11 11:06:56 +01:00
Patrick von Platen ff648221bd fix conflicts 2020-03-11 11:06:56 +01:00
Patrick von Platen c0d9dd3ba9 refactored code a bit and made more generic 2020-03-11 11:06:56 +01:00
Patrick von Platen d8e2b3c547 fix conflicts 2020-03-11 11:06:56 +01:00
Sam Shleifer ed37f9fa4f [Bart] _prepare_decoder_inputs should use large negative (#3158) 2020-03-06 16:06:36 -05:00
patrickvonplaten 58fc8f97a3 fix renaming problem 2020-03-06 00:35:47 +01:00
Sam Shleifer 857e0a0d3b Rename BartForMaskedLM -> BartForConditionalGeneration (#3114)
* improved documentation
2020-03-05 17:41:18 -05:00
sshleifer 1360dacaa3 cleanup deltas 2020-03-05 12:57:42 -05:00
sshleifer c36fdc88d4 tests pass 2020-03-05 12:33:08 -05:00
Sam Shleifer e9e6efdc45 BartForSequenceClassification: fix num_labels, add test (#3110) 2020-03-03 15:54:29 -05:00
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
Julien Chaumond f5516805c2 Fix bart slow test 2020-02-26 20:47:49 +00:00
Julien ChaumondandSam Shleifer 9cda3620b6 Fix (non-slow) tests on GPU (torch) (#3024)
* Fix tests on GPU (torch)

* Fix bart slow tests

Co-authored-by: Sam Shleifer <sshleifer@gmail.com>
2020-02-26 11:59:25 -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 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