Skip to content

Add mlflow support for generate_during_eval DPOTrainer #3660

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom
Merged
merged 9 commits into from
Jul 2, 2025
Merged

Add mlflow support for generate_during_eval DPOTrainer #3660

merged 9 commits into from
Jul 2, 2025

Conversation

dhruvmullick
Copy link
Contributor

@dhruvmullick dhruvmullick commented Jun 27, 2025 •
edited by shirinyamani
Loading

What does this PR do?

Addresses #3659
Will allow logging a table with eval generations when generate_during_eval is enabled for the DPO Trainer

Tested with a DPO run

Who can review?

@qgallouedec

Minimal script for testing;

from datasets import load_dataset
from trl import DPOConfig, DPOTrainer
from transformers import AutoModelForCausalLM, AutoTokenizer

model = AutoModelForCausalLM.from_pretrained("Qwen/Qwen2-0.5B-Instruct")
tokenizer = AutoTokenizer.from_pretrained("Qwen/Qwen2-0.5B-Instruct")
train_dataset = load_dataset("trl-lib/ultrafeedback_binarized", split="train[:100]")

training_args = DPOConfig(output_dir="Qwen2-0.5B-DPO", generate_during_eval=True, logging_steps=1, report_to="mlflow")

trainer = DPOTrainer(model=model, args=training_args, processing_class=tokenizer, train_dataset=train_dataset)
trainer.train()

then running by;

accelerate launch train.py

@dhruvmullick dhruvmullick mentioned this pull request Jun 27, 2025
Closed
@dhruvmullick dhruvmullick marked this pull request as ready for review June 27, 2025 22:32
@dhruvmullick dhruvmullick mentioned this pull request Jun 30, 2025
Merged
@HuggingFaceDocBuilderDev
Copy link

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.

@shirinyamani
Copy link
Member

shirinyamani commented Jul 2, 2025 •
edited
Loading

Hi @dhruvmullick thank, and sure if you need the mlflow logging! I've added minimal script on how to test it below;

@shirinyamani shirinyamani self-requested a review July 2, 2025 20:39
shirinyamani
shirinyamani reviewed Jul 2, 2025
View reviewed changes
Copy link
Member

@shirinyamani shirinyamani left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

adding the scrpit to the description for future reference

@shirinyamani shirinyamani self-requested a review July 2, 2025 20:40
@shirinyamani shirinyamani merged commit 90cdf96 into huggingface:main Jul 2, 2025
9 of 10 checks passed
@dhruvmullick dhruvmullick deleted the add-mlflow-dpotrainer branch July 2, 2025 20:47
@dhruvmullick
Copy link
Contributor Author

Awesome, thanks for the review and the fixes! @shirinyamani

marcandrelarochelle pushed a commit to marcandrelarochelle/trl that referenced this pull request Jul 29, 2025
@dhruvmullick @shirinyamani @marcandrelarochelle
🖼️ Add mlflow support for generate_during_eval DPOTrainer (huggingfac…
ad6453e 
…e#3660)

Co-authored-by: Shirin Yamani <75791599+shirinyamani@users.noreply.github.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@shirinyamani shirinyamani shirinyamani approved these changes

Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
3 participants
@dhruvmullick @HuggingFaceDocBuilderDev @shirinyamani