Fix FileNotFoundError in DPO/KTO precompute_ref_log_probs - #6348
Merged
Conversation
In _precompute_ref_logps(), dataset.map() was called without cache_file_name, causing the datasets library to write cache to a different path than the one TRL computed and subsequently tried to read from. Pass cache_file_name=cache_file so the cache is written to the expected location. Fixes KTOTrainer._precompute_ref_logps() which had the same bug. Fixes #6291 Test Plan: python -c " from datasets import Dataset from trl import DPOConfig, DPOTrainer from transformers import AutoModelForCausalLM, AutoTokenizer model_name = 'danish-foundation-models/munin-apertus-8b' tokenizer = AutoTokenizer.from_pretrained(model_name) tokenizer.pad_token = tokenizer.eos_token model = AutoModelForCausalLM.from_pretrained(model_name, device_map='cpu') ds = Dataset.from_dict({'prompt': ['x']*4, 'chosen': ['a']*4, 'rejected': ['b']*4}) cfg = DPOConfig(output_dir='/tmp/test-dpo', precompute_ref_log_probs=True) trainer = DPOTrainer(model=model, ref_model=None, args=cfg, train_dataset=ds, processing_class=tokenizer) print('DPOTrainer initialized successfully') " # Expected: initializes without FileNotFoundError Signed-off-by: Yuchen Fan <functionhx@gmail.com>
Contributor
|
Your PR has been closed because it does not follow the required PR template. Please open a new PR following the template. See CONTRIBUTING.md for guidance. |
Member
|
Thanks, I'll approve this pr exceptionally because the fix is correct and small, but please in the future remember that we don't welcome purely AI generated PRs |
qgallouedec
approved these changes
Jul 9, 2026
|
The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update. |
Contributor
Author
|
Noted, thanks. |
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.
What does this PR do?
Fixes #6291. When precompute_ref_log_probs=True, the computed cache_file path wasn't passed to dataset.map(), so datasets writes to a different random path than TRL reads from -- causing FileNotFoundError. Fixed in both DPOTrainer and KTOTrainer.
Before submitting
precompute_ref_log_probs=True#6291AI writing disclosure
Who can review?
Anyone in the community.