FIX: TRL trainer preprocessing step was running in one process #1583
Conversation
This file contains 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
…ainer args and directly to the trainer when DPO
|
I made a misstake, i sent it as trainer args in all cases, but it only applies CPO, KTO and ORPO trainer args. For the DPO it needs to be sent directly to the trainer. My latest commit in the PR fixes that. |
djsaunde
pushed a commit
that referenced
this pull request
Dec 17, 2024
* FIX: TRL trainer preprocessing step was running in one process * FIX: Changed so that dataset_num_proc is sent to CPO, KTO and ORPO trainer args and directly to the trainer when DPO * FIX: Changed back to only support ORPO for now, since KTO is handled in another way --------- Co-authored-by: Ali Mosavian <ali.mosavian@kry.se>
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Outdated suggestions cannot be applied.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.
Description
We weren't passing dataset_num_proc to TRL training config, thus the initial data preprocessing steps in the TRL trainer was running in one process only.
Motivation and Context
Speeds up training start time by a lot depending on the number logical cores you have.
How has this been tested?
Tested it with the ORPO trainer in axolotl.
Screenshots (if appropriate)
Types of changes
Social Handles (Optional)
https://www.linkedin.com/in/ali-mosavian-7a27457/
https://github.com/ali-mosavian