Small updates for compatibility with WebDataset pipelines #171
Conversation
This file contains hidden or 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
✅ Result of Pytest Coverage---------- coverage: platform linux, python 3.10.0-final-0 -----------
|
kerimoglutolga
force-pushed
the
tolga/WebDatasetUpdates
branch
from
6af7f96 to
cc55aa4
Compare
MaxiBoether
requested changes
Apr 4, 2025
Contributor
MaxiBoether
left a comment
MaxiBoether
left a comment
There was a problem hiding this comment.
Thanks for the work! Let's clean this up a bit
Comment on lines
-2
to
+6
| import os | ||
| import argparse | ||
| import requests | ||
| import os | ||
| from concurrent.futures import ThreadPoolExecutor, as_completed | ||
|
|
||
| import requests |
Contributor
There was a problem hiding this comment.
Unnecessary changes, no?
| "CC12MDataset", | ||
| "MSCOCODataset", | ||
| "LAION400MDataset", | ||
| "COYO700MDataset", |
Contributor
There was a problem hiding this comment.
As discussed, this must change
Comment on lines
+9
to
+30
| class MixteraDataPipeline(wds.DataPipeline): | ||
| """ | ||
| Supports building arbitrary webdataset pipelines with Mixtera's `MixteraTorchDataset` as the data source. | ||
| """ | ||
|
|
||
| def __init__( | ||
| self, | ||
| client: MixteraClient, | ||
| query: Query, | ||
| query_execution_args: QueryExecutionArgs, | ||
| result_streaming_args: ResultStreamingArgs, | ||
| pipeline: Iterable[Any], | ||
| ): | ||
| super().__init__(*pipeline) | ||
| self.client = client | ||
| self.query = query | ||
| self.query_execution_args = query_execution_args | ||
| self.result_streaming_args = result_streaming_args | ||
|
|
||
| torch_dataset = MixteraTorchDataset( | ||
| client=client, | ||
| query=query, |
Contributor
There was a problem hiding this comment.
I don't understand what this is necessary for. The MixteraTorchDataset is a user facing abstraction
Comment on lines
-10
to
+9
| def decode(sample: dict[str, Any], decode_image: bool = True) -> dict[str, Any]: | ||
| def decode_sample(sample: dict[str, Any]) -> dict[str, Any]: |
Contributor
There was a problem hiding this comment.
It's weird to have a funciton called decode_sample which you can optionally tell to not decode. Can you find a cleaner way to disable image decoding?
Comment on lines
+56
to
+64
| class MMIndexedTarRawBytes(MMIndexedTar): | ||
| """ | ||
| A subclass of `MMIndexedTar` that returns the raw bytes instead of an IOBytes object. | ||
| """ | ||
|
|
||
| def get_file(self, i: int) -> tuple[str, bytes]: | ||
| filename, data = self.get_at_index(i) | ||
| return filename, data | ||
|
|
| return self.decoder(sample) | ||
| raise ValueError("Co") | ||
| sample[ext[1:]] = data | ||
| sample["__key__"] = key # type: ignore |
| self.add_metadata(sample_id=line_number, dataset=dataset_name) | ||
|
|
||
|
|
||
| class DomainNetMetadataParser(MetadataParser): |
Contributor
There was a problem hiding this comment.
We need to clean up the datasets vs metadata parsers here. Not sure why in addition to the datasets you added you also added this parser. Let's find a cleaner abstraction for this. You could have like a "LlavaMetadataParser" that somehow e.g. from the path infers which dataset the sample comes from. Or you have one MetadataParser per dataset. You can actually register multiple datasets within the same data collection so that should also not be an issue.
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.
No description provided.