mehrad
|
69c0d5f801
|
Fix and update tests
|
2020-12-03 17:24:55 -08:00 |
mehrad
|
8b644cca02
|
bump transformers version to 3.5.1
|
2020-12-03 12:50:34 -08:00 |
s-jse
|
653424214c
|
Merge pull request #58 from stanford-oval/wip/lower-memory
Submission work
|
2020-12-03 00:23:03 -08:00 |
mehrad
|
f0a77fc172
|
Fix and add test for paired encoding
|
2020-12-02 21:55:35 -08:00 |
mehrad
|
2384ae9b1e
|
Use bart-tiny-random instead of bart-large for tests
|
2020-11-28 14:57:34 -08:00 |
mehrad
|
c3bc5835ad
|
Add translation test files
|
2020-11-15 14:33:45 -08:00 |
mehrad
|
dbfdfb20d3
|
run_lm_finetuning: fix attention_mask
|
2020-11-15 14:22:28 -08:00 |
mehrad
|
744a7a8483
|
run_generation.py: fix and update denoising functions
|
2020-11-15 13:46:22 -08:00 |
mehrad
|
b84a6548a6
|
Changes to make run_lm_finetuning compatible with mbart and marian
|
2020-11-15 13:46:17 -08:00 |
Sina
|
2f94fc2d35
|
Fix bug
|
2020-11-11 17:21:12 -08:00 |
Sina
|
e2ad021972
|
Clarify test case
|
2020-11-07 23:42:27 -08:00 |
Sina
|
9a847b4b1c
|
Add Bart test
|
2020-11-07 15:19:44 -08:00 |
Sina
|
cdcfa97ce0
|
Fix bugs
|
2020-07-24 16:32:54 -07:00 |
Sina
|
44021e2ba0
|
Add test for server mode
|
2020-07-24 16:32:54 -07:00 |
mehrad
|
5c0453a4ec
|
Add multilingual dialog almond tasks
|
2020-07-22 22:04:27 -07:00 |
mehrad
|
813e28b8fa
|
Update tests
|
2020-07-20 22:49:31 -07:00 |
Sina
|
a943b2a1b5
|
Postpone BART tests until a later transformers version
|
2020-07-03 18:46:29 -07:00 |
Sina
|
04104f0dc3
|
Free up space during tests
|
2020-07-03 17:59:10 -07:00 |
Sina
|
f8976ac6d7
|
Tests are less resource-intensive to run on Travis
|
2020-07-03 17:14:24 -07:00 |
Sina
|
31fca52dde
|
Fix BART tests
|
2020-07-03 16:44:29 -07:00 |
Sina
|
eedb4b6b5e
|
Merge branch 'master' into paraphrase-merge
|
2020-07-03 15:09:51 -07:00 |
mehrad
|
6de6bd98a5
|
Added option to use language in the question
|
2020-06-23 14:58:43 -07:00 |
mehrad
|
0489d19f2e
|
Option to pool encoded values by choosing cls token representation
|
2020-06-23 14:58:43 -07:00 |
mehrad
|
a93dfc9099
|
Merge BART code with the rest of paraphrasing codebase
|
2020-05-04 15:54:09 -07:00 |
mehrad
|
50c4de9d60
|
merge master branch
|
2020-05-04 10:06:35 -07:00 |
mehrad
|
8ba1f4c490
|
code refactoring
|
2020-05-04 10:03:08 -07:00 |
mehrad
|
5f871512a6
|
bart_evaluation is integrated with run_generation script
|
2020-04-29 21:14:03 -07:00 |
mehrad
|
9d5945376f
|
move files to test dir
|
2020-04-27 16:36:36 -07:00 |
mehrad
|
2eb78d7080
|
add tests
|
2020-04-27 16:30:19 -07:00 |
mehrad
|
36b0fb5317
|
fix bug in iter
|
2020-04-19 18:03:11 -07:00 |
Sina
|
e24e426a3c
|
test datasets should not be ignored
|
2020-04-18 19:54:48 -07:00 |
Sina
|
a122c05062
|
paraphrase training is now done with tsv files
|
2020-04-18 18:46:45 -07:00 |
mehrad
|
891bee8632
|
merge master branch
|
2020-04-09 22:10:35 -07:00 |
mehrad
|
31190c3f18
|
update tests
|
2020-04-06 16:03:05 -07:00 |
mehrad
|
f3bf392231
|
adding tests + fixing bugs
|
2020-03-30 14:43:03 -07:00 |
mehrad
|
703edc4fdc
|
Option to do sentence batching
Batches can have multiple minibatches; each containing same sentences but in different languages
|
2020-03-30 12:50:02 -07:00 |
mehrad
|
60adad8173
|
adding tests + bug fixes
|
2020-03-25 23:51:04 -07:00 |
mehrad
|
a2d2e740de
|
adding multilingual task
|
2020-03-25 02:33:19 -07:00 |
Sina
|
06131f12dc
|
Merge branch 'master' into wip/sinaj/clean
|
2020-03-02 22:27:15 -08:00 |
Sina
|
cc258c0e2a
|
Fixed beam search
|
2020-03-02 22:06:03 -08:00 |
Sina
|
ea078d8e46
|
Added tests and instructions for paraphrasing
|
2020-03-02 16:57:35 -08:00 |
mehrad
|
843913c951
|
cleanup
|
2020-03-02 11:23:56 -08:00 |
mehrad
|
daa7131675
|
fix decoding
|
2020-02-26 02:07:31 -08:00 |
Giovanni Campagna
|
ed10e1b916
|
Rename the python package to genienlp
|
2020-01-28 18:06:50 -08:00 |
Giovanni Campagna
|
2762720bcb
|
Add a BiLSTM encoder
To produce a meaningful initial state for the BiLSTM on the decoder,
and to further encode the output of BERT
|
2020-01-17 21:41:31 -08:00 |
Giovanni Campagna
|
92ac4003e7
|
Allow combining BERT (Identity) encoder with LSTM decoder
Make up a zero initial state of the RNN, and use attention.
This should improve the convergence of the decoder and at least
learning valid syntax.
|
2020-01-17 20:30:44 -08:00 |
Giovanni Campagna
|
1f7edc7b24
|
BERT
|
2020-01-16 15:34:16 -08:00 |
Giovanni Campagna
|
d569037078
|
Make tests pass
Redo how predict works, fix reversing
|
2020-01-14 09:26:33 -08:00 |
mehrad
|
3a1b969ad3
|
updates
|
2019-03-01 17:51:54 -08:00 |
mehrad
|
9c3145846a
|
add tests
|
2019-03-01 17:51:54 -08:00 |