abc65eaf53
Remove `train-paraphrase` command
2022-08-26 12:04:35 -07:00
bc82aeb4b6
Add test to check the outputs of --is_hf_model
2022-08-25 20:29:37 -07:00
c046ec0eb0
Fix tests
2022-08-25 18:37:49 -07:00
25ebee3964
Factor out some shared parameters in test cases
2022-08-24 21:20:05 -07:00
c5d04f31bc
Make tests multi-line commands
2022-08-24 21:04:08 -07:00
16b7c7a717
Add test for direct HF model prediction
2022-08-24 20:40:48 -07:00
996b406184
Add test for bitod_error_cls task
2022-08-08 11:56:31 -07:00
c1c38298ec
Update OOD expected results
2022-08-05 14:20:33 -07:00
101a6d3f6c
Update e2e tests and results
2022-06-29 14:35:56 -07:00
6e8e7f1f84
Update e2e_dialogue_preds.json
...
Due toX_re changes in bitod/main (7b1e6ae9c4fdfc124140ca33ff71626037a39820)
2022-06-16 16:59:11 -07:00
968b708336
Fix and update tests
2022-05-09 14:16:41 -07:00
7ce4cdaf7c
Add test for entity translation
2022-05-09 10:48:12 -07:00
1613516699
build(deps): bump transformers from 4.16.2 to 4.17.0 ( #256 )
...
* build(deps): bump transformers from 4.16.2 to 4.17.0
Bumps [transformers](https://github.com/huggingface/transformers ) from 4.16.2 to 4.17.0.
- [Release notes](https://github.com/huggingface/transformers/releases )
- [Commits](https://github.com/huggingface/transformers/compare/v4.16.2...v4.17.0 )
---
updated-dependencies:
- dependency-name: transformers
dependency-type: direct:production
update-type: version-update:semver-minor
...
Signed-off-by: dependabot[bot] <support@github.com>
* Bart tokenizer no longer inherits from Roberta
* Add XGLMTokenizer
* Update TokenClassification test results
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
Co-authored-by: mehrad <mehrad@stanford.edu>
2022-03-07 11:14:59 -08:00
4fe3f06bed
test_paraphrasing: compute rouge1 and rougeL
2022-03-02 17:39:44 -08:00
9a6de26afb
Update e2e tests
...
Use a pretrained marian model instead of randomly initialized bart so we get some non-trivial results.
Shrink bitod test datasets. Now that we are testing e2e_dialogue_evaluation it will help to keep testing time almost the same.
2022-03-02 16:42:26 -08:00
26fb840339
Test evaluate-file
2022-03-02 16:27:22 -08:00
941881fb59
test_e2e_dialogues: test e2e prediction too
2022-03-02 16:27:21 -08:00
1b790f6658
Remove test_main_almond_multilingual
2022-02-28 13:22:45 -08:00
96f6db7e3f
Remove data caching
...
it's almost never been used
2022-02-28 13:19:39 -08:00
e07c8c5216
Update test results after adding input column to predictions
2022-02-25 14:01:41 -08:00
86044a8d9b
Add tests for bitod tasks
2022-02-25 13:33:27 -08:00
53746f351e
Fix prediction with `num_beams > 1` and `num_outputs > 1`
2022-02-21 20:51:36 -08:00
633ab40ce8
Update tests
2021-12-15 15:25:31 -08:00
1d50ba072e
Fix bug in server for TransformerLSTM ( #229 )
2021-11-29 10:51:47 -08:00
699b784483
predict: assing eval_lang to pred_lang if not provided
2021-10-06 17:33:41 -07:00
302075f32c
Update bleu test expected results
...
In sacrebleu 2.0 except bleu score is 0.0 if there are no ngram matches (i.e. smoothing is skipped)
2021-09-20 09:53:25 -07:00
b0e02a72d7
Address PR comments
...
- change default names for add_entities_to_text and ned_normalize_types
- improve code style and comments
- move banned_phrases to a text file
2021-09-06 17:17:26 -07:00
727e607040
Remove --database_lookup_method argument
...
and some cleanups
2021-09-06 17:17:24 -07:00
9dc6295d45
Use marisa_trie for alias2qids and qid2typeqid to save memory
2021-09-06 17:17:24 -07:00
b64b952948
Rename bootleg_post_process_types to ned_normalized_types
...
- the argument is not specific to bootleg anymore and can be used with Naive and Oracle too
- it now hast three modes: no, yes, and force. previous functionlity was doing only yes and no.
2021-09-06 17:17:24 -07:00
911fef976c
Add a test script for cuda errors
2021-08-10 03:42:08 -07:00
e584ac70d4
Rename canonical2type.json to alias2type.json
2021-08-05 01:16:56 -07:00
d1bc56427e
Merge pull request #183 from stanford-oval/wip/mehrad/pr_10
...
NED Refactoring (2)
2021-08-04 10:28:50 -07:00
13b8f23f14
Address PR comments (2)
2021-08-02 13:07:40 -07:00
df01c76924
Remove ElasticDatabase and elasticsearch dep
2021-08-01 20:55:33 -07:00
b46d477724
Address PR comments
2021-08-01 20:55:32 -07:00
49ab643afd
Updates to make tests pass
2021-08-01 12:10:45 -07:00
bcb5ea5528
Remove nonessential files from database dir
2021-08-01 12:10:44 -07:00
455dc7c869
Use add_entity_to_text and entity_attributes
...
add_entity_to_text tells how to add entity attributes to text.
entity_attributes indicates what to add (e.g. type, qid)
2021-08-01 12:09:43 -07:00
b52ac345e4
Rename Feature to Entity
2021-08-01 12:09:42 -07:00
f404fcc130
Add prediction and compare expected results in ood test script
2021-08-01 02:35:22 -07:00
0a5b3a6f4b
Fix dataloader and dataset
2021-07-31 05:18:32 -07:00
9f6c414aee
Split training and test data
2021-07-29 02:46:51 -07:00
61cabfa07f
Fix bugs in generate_with_classification_model
2021-07-27 08:13:01 -07:00
6837d40a43
Add ood dataloader, model, and toy dataset
2021-07-27 02:29:43 -07:00
c4bad249eb
Major refactoring of alignment code
2021-07-23 17:47:22 -07:00
3d9708b3cb
predict: include gold answers in prediction file
...
- so running local evaluation is possible without accessing the original dataset
- easier debugging since answer might have been changes after passing through genienlp
2021-07-23 17:47:13 -07:00
871e5d059c
Updates to make NED tests pass
2021-07-14 11:45:13 -07:00
724dc23b57
Simplify how we keep track of spans
2021-06-29 21:45:24 -07:00
d96ddcd08f
Misc. code updates
...
update translation tests, address some bugs
2021-06-29 19:41:39 -07:00
Sina