support reward modeling and ppo #2093
Merged
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
Jintao-Huang
approved these changes
Sep 23, 2024
hjh0119
changed the title
hjh0119
commented
Oct 8, 2024
| @@ -98,7 +99,7 @@ def __init__(self, | |||
| optimizers=optimizers, | |||
| preprocess_logits_for_metrics=preprocess_logits_for_metrics, | |||
| **kwargs) | |||
| if not self.label_names: | |||
| if not hasattr(self, 'label_names') or not self.label_names: | |||
There was a problem hiding this comment.
PPOv2Trainer does not have the label_names attribute.
hjh0119
changed the title
hjh0119
commented
Oct 9, 2024
| @@ -98,7 +98,8 @@ def prepare_model(model, args: SftArguments): | |||
| if args.resume_from_checkpoint is None: | |||
| handle_target_modules(model, args) | |||
| handle_modules_to_save(model, args) | |||
| if args.init_lora_weights and args.init_lora_weights.lower() in ('true', 'false'): | |||
| if args.init_lora_weights and isinstance(args.init_lora_weights, | |||
| str) and args.init_lora_weights.lower() in ('true', 'false'): | |||
There was a problem hiding this comment.
Avoid errors in reward_model_args.
hjh0119
commented
Oct 9, 2024
| @@ -1196,6 +1196,10 @@ def _init_training_args(self) -> None: | |||
| if 'accelerator_config' in parameters: | |||
| kwargs['accelerator_config'] = {'dispatch_batches': False} | |||
|
|
|||
| metric_for_best_model = 'rouge-l' if self.predict_with_generate else 'loss' | |||
| if hasattr(self, 'rlhf_type') and self.rlhf_type == 'ppo': | |||
| metric_for_best_model = None | |||
There was a problem hiding this comment.
In the PPO training metrics, there are no metrics that start with "eval", set None here
Jintao-Huang
approved these changes
Oct 9, 2024
tastelikefeet
approved these changes
Oct 10, 2024
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.
PR type
PR information
support
reward modelingfor LLM and MLLMsupport
PPOfor LLM