feat: GPTDatasetFolder for easier definition of dataset-based mixtures#58
Open
MaxiBoether wants to merge 6 commits into
Open
feat: GPTDatasetFolder for easier definition of dataset-based mixtures#58MaxiBoether wants to merge 6 commits into
GPTDatasetFolder for easier definition of dataset-based mixtures#58MaxiBoether wants to merge 6 commits into
Conversation
MaxiBoether
changed the title
GPTDatasetFolder for easier definition of dataset-based mixtures
MaxiBoether
commented
Mar 11, 2025
Comment on lines
+460
to
+476
| for i, _split in enumerate(Split): | ||
| if split[i] is None: | ||
| mid_level_datasets.append(None) | ||
| else: | ||
| mid_level_datasets.append( | ||
| self.build_generic_dataset( | ||
| self.cls, | ||
| self.is_built_on_rank, | ||
| synchronize_ranks, | ||
| None, # indexed_dataset (unused) | ||
| dataset_path, # folder_path | ||
| None, # indexed_indices (unused) | ||
| sizes[i], | ||
| _split, | ||
| self.config, | ||
| ) | ||
| ) |
Author
There was a problem hiding this comment.
I know we're not really using splits in our runs but it was weird not respecting it. I am not 100% sure this would actually work if you used splits though.
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
This PR adds a middle layer in after the
GPTDataset(per file prefix) to accomodate the common setting where we have one directory per dataset with multiple files for this dataset in that directory. Then, theGPTDatasetFoldercan be merged using the existingBlendedDataset.It can be used with our existing runs sbatch files. The only thing that needs to change is instead of calling
MEGATRON_LM_DIR/scripts/tools/create_data_config.pyin the DATA_ARGS, you just need to pass a space separated list of directories (and optionally, a weight after each directory).TODO: Implement looping on each directory (@TJ-Solergibert is on that) by moving looping to the BlendedDataset.