Closes #629 | Add/Update Dataloader VLSP2020 MT

[patrickamadeus]

Closes #629

Checkbox

• Confirm that this PR is linked to the dataset issue.
• Create the dataloader script seacrowd/sea_datasets/{my_dataset}/{my_dataset}.py (please use only lowercase and underscore for dataset folder naming, as mentioned in dataset issue) and its __init__.py within {my_dataset} folder.
• Provide values for the _CITATION, _DATASETNAME, _DESCRIPTION, _HOMEPAGE, _LICENSE, _LOCAL, _URLs, _SUPPORTED_TASKS, _SOURCE_VERSION, and _SEACROWD_VERSION variables.
• Implement _info(), _split_generators() and _generate_examples() in dataloader script.
• Make sure that the BUILDER_CONFIGS class attribute is a list with at least one SEACrowdConfig for the source schema and one for a seacrowd schema.
• Confirm dataloader script works with datasets.load_dataset function.
• Confirm that your dataloader script passes the test suite run with python -m tests.test_seacrowd seacrowd/sea_datasets/<my_dataset>/<my_dataset>.py or python -m tests.test_seacrowd seacrowd/sea_datasets/<my_dataset>/<my_dataset>.py --subset_id {subset_name_without_source_or_seacrowd_suffix}.
• [.] If my dataset is local, I have provided an output of the unit-tests in the PR (please copy paste). This is OPTIONAL for public datasets, as we can test these without access to the data files.

Tests

indomain-news (3 splits)

basic (1 split)

VLSP20-official (1 split)

NB

• Skipping openSub & mono-vi for future development (Large Drive file download bottleneck), already opened thread in Discord discussion.
• Tested 3 out of 7 subsets

[sabilmakbar]

As of now, the dataloader and its metadata work fine. However, since there might be some issue with the data split definition, I'll revisit it once it has been addressed.

[patrickamadeus]

It's done! @sabilmakbar thankyou for the review ☺️

[sabilmakbar]

By the way, @patrickamadeus, did you happen to observe the different number of samples generated from your dataloader implementation compared to the one reported in their GH?

I noticed two subsets have minor amt of sample diff:

1. Subset VLSP20-official 789 (reported on Src GH) vs 790 (generated examples)
2. wiki-alt 20000 (reported on Src GH) vs 20106 (generated examples)

[patrickamadeus]

By the way, @patrickamadeus, did you happen to observe the different number of samples generated from your dataloader implementation compared to the one reported in their GH?

I noticed two subsets have minor amt of sample diff:

1. Subset VLSP20-official 789 (reported on Src GH) vs 790 (generated examples)
2. wiki-alt 20000 (reported on Src GH) vs 20106 (generated examples)

Hi @sabilmakbar ! That's interesting, they mentioned that they have N sentences. Actually I splitted the dataset every \n instead of assuming . for the end of the sentence (would even explode the numbers even more).

Since they don't give any clear definition of how to define each "sentence", do you think it's reasonable for now if we assume each line for each sentence? Because I have reviewed each dataset and it's true that each generated example's number matches to each file's line count.

[sabilmakbar]

Hi @sabilmakbar ! That's interesting, they mentioned that they have N sentences. Actually I splitted the dataset every \n instead of assuming . for the end of the sentence (would even explode the numbers even more).

Since they don't give any clear definition of how to define each "sentence", do you think it's reasonable for now if we assume each line for each sentence? Because I have reviewed each dataset and it's true that each generated example's number matches to each file's line count.

Okay then, since I think everything else has the same number as reported, we can acknowledge it (prob adding it to inline comment would be better).

[sabilmakbar]

thanks for the work, @patrickamadeus! let's wait for @raileymontalan's review

[holylovenia]

Hi @raileymontalan, I would like to let you know that we plan to finalize the calculation of the open contributions (e.g., dataloader implementations) in 31 hours, so it'd be great if we could wrap up the reviewing and merge this PR before then.

cc: @patrickamadeus

[fhudi]

Sorry for directly changing the files to follow the black formatter.
Normally I would just request for changes, but it already passed the test without formatter hence I directly changed it.

LGTM.
All subsets passed the test.
Passed the reviewing checklist.
Thanks for the hard-work.
Merging this soon.

---

cc: @sabilmakbar