refactor: Split dataset into 'StaticDataset' and 'GenerativeDataset' #1801
+977
−713
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
initialized from HF/Pytorch/JSON/list of Dicts, remove the need for setup call and subsequently cleanup
test coverage
instead of strings and add seed for reproducibility
between simply skipping invalid datapoints in a seed dataset and throwing an exception
seed dataset to ensure they are defined before the other functions are
to store data in Seed Dataset
changes in base.py
requirements
getitem and cast len(data) to a Sized to pass mypy tests
…of Seed Dataset to ensure proper validation & add additional logging to init from JSON
call helper functions for each type of data
initialising with PyTorch Datasets
properly cover all 4 initialization functions & add tests for sampling
|
TO DO:
|
This was referenced Mar 11, 2025
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Description
This PR simplifies the current approach, where we define
BaseDataset,SeedDataset,SyntheticDatasetandGenerativeDataset. Specifically, it combinesBaseDataset,SeedDatasetandSyntheticDatasetinto one class calledStaticDataset, as they are all static, i.e. do not change size at runtime.GenerativeDataseton the other hand generates new samples at runtime and is hence structurally different, thus deserving its own class. This PR adaptsGenerativeDatasetto now take aStaticDatasetas attribute (before, it took aSeedDataset), which is used to generate new synthetic datapoints. This PR also adds a way to save the data generated at runtime into a JSONL file (such that it can be loaded efficiently for downstream purposes).Lastly, this PR removes a lot of code that is unused and whose functionality has been implemented elsewhere.
Checklist
Go over all the following points, and put an
xin all the boxes that apply.Fixes #issue-numberin the PR description (required)pyproject.tomlanduv lockIf you are unsure about any of these, don't hesitate to ask. We are here to help!
Note to reviewers
This PR builds on top of refactor Seed Dataset to improve compatibility and simplify usage #1734.
GenerativeDatasetshould be anIterableDataset, not aDataset. This will be addressed in another PR, once this is merged. ([Refactor] RefactorGenerativeDatasetfor Efficient Synthetic Data Storage and Retrieval #1728)