[feature] Add debug_dataloader_samples utility to preview decoded dataloader samples (#184) #364
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
…face#353) * can only merge to main from dev * Fix UnBoundLocalError in `clm_collator.py` (huggingface#339) * Update clm_collator.py * can only merge to main from dev (huggingface#348) --------- Co-authored-by: Nouamane Tazi <[email protected]> * fix init and init scaling factor and run evals in background (huggingface#349) * InitScalingMethod * InitScalingMethod * run evals in background (huggingface#352) * eval * try adding lightevalrunner to trainer * amend * amend * amend * amend * amend * amend * . * amend * amend * . * qos to low * add nanotron_path * some fix: logs, and config * cp instead of sync * eval_interval * serialize sanity checks * add output dir and s3_save path in the config * fix s3 only if define * fixes --------- Co-authored-by: elie <[email protected]> Co-authored-by: “eliebak” <[email protected]> --------- Co-authored-by: elie <[email protected]> Co-authored-by: “eliebak” <[email protected]> --------- Co-authored-by: Connector Switch <[email protected]> Co-authored-by: elie <[email protected]> Co-authored-by: “eliebak” <[email protected]>
* deepwiki * Update README.md
NouamaneTazi
requested changes
May 15, 2025
Comment on lines
+152
to
+154
| _fused_rotary_emb: bool = False | ||
| _fused_rms_norm: bool = False | ||
| _use_qkv_packed: bool = False |
Comment on lines
+65
to
+70
| slurm_job_id, slurm_log = run_slurm_one_job( | ||
| config=self.config, | ||
| lighteval_config=self.lighteval_config, | ||
| model_checkpoint_path=checkpoint_path, | ||
| current_step=self.config.general.step, | ||
| ) |
|
Closing this in facor of a cleaner PR: #368 |
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.
What does this PR do?
This PR adds a utility function
debug_dataloader_samples()underutils/debug_utils.py.It allows users to preview a few decoded examples from the first batch of the training
DataLoader, which can help ensure that tokenization is working correctly before launching a long training run.This debug utility addresses #184.
Although another contributor expressed interest in the issue, no progress or maintainer feedback has occurred in the last 3 weeks.
As this feature is simple but useful, I decided to implement and contribute it directly.
How to use
This utility can be optionally called from
run_train.pyor any other training script to inspect tokenized input samples before training.Here is an example snippet:
Notes
To avoid a circular import issue between
utils.pyanddistributed.py,this function is placed in a new module
debug_utils.pyinstead of the mainutils.py.Logging is used instead of
print()to remain consistent with Nanotron’s logging system.Fixes #184
Before submitting
Who can review?
Anyone in the community is free to review the PR once the tests have passed. Feel free to tag
members/contributors who may be interested in your PR.