[recipe] feat: integrate DAPO and provide reproduction script#623
Merged
[recipe] feat: integrate DAPO and provide reproduction script#623
Conversation
Co-authored-by: Guangming Sheng <petershengwhu@gmail.com>
…e for low and high (#618) - As titled
PeterSH6
reviewed
Mar 16, 2025
Closed
PeterSH6
reviewed
Mar 31, 2025
Collaborator
|
LGTM! |
PeterSH6
approved these changes
Apr 3, 2025
verl/trainer/ppo/ray_trainer.py
Outdated
| for sample_idx, data_source in enumerate(data_sources): | ||
| prompt = sample_inputs[sample_idx] | ||
|
|
||
| var2vals = data_src2prompt2var2vals[data_source][prompt] |
Collaborator
There was a problem hiding this comment.
What's the meaning of data_src2prompt2var2vals
Collaborator
Author
There was a problem hiding this comment.
2meansto, constructing the multi-level dictvarmeans a kind of "variable" such asacc,final_rewardvalsmeans the "values" of thevariables of the trajectories after somepromptfrom somedata_src
I will add comments in a future PR to make it clearer.
Collaborator
|
Really nice job!!! |
3 tasks
yushengsu-thu
pushed a commit
to yushengsu-thu/verl
that referenced
this pull request
Apr 4, 2025
…roject#623) > [!WARNING] > As mentioned in verl-project#623 (comment), the implementation of gradient accumulation in verl has been only compatible with the sequence-mean loss, but all the DAPO experiments with the token-mean loss were run with the incompatible implementation. > **We keep it as is for reproducibility in this branch** and will fix it in another PR for the main branch. --------- Co-authored-by: Guangming Sheng <shengguangming@bytedance.com> Co-authored-by: Guangming Sheng <petershengwhu@gmail.com>
yuchenwang3
pushed a commit
to yuchenwang3/verl
that referenced
this pull request
Apr 25, 2025
…roject#623) > [!WARNING] > As mentioned in verl-project#623 (comment), the implementation of gradient accumulation in verl has been only compatible with the sequence-mean loss, but all the DAPO experiments with the token-mean loss were run with the incompatible implementation. > **We keep it as is for reproducibility in this branch** and will fix it in another PR for the main branch. --------- Co-authored-by: Guangming Sheng <shengguangming@bytedance.com> Co-authored-by: Guangming Sheng <petershengwhu@gmail.com>
histmeisah
pushed a commit
to SJTU-IAAR/verl
that referenced
this pull request
Apr 27, 2025
…roject#623) > [!WARNING] > As mentioned in verl-project#623 (comment), the implementation of gradient accumulation in verl has been only compatible with the sequence-mean loss, but all the DAPO experiments with the token-mean loss were run with the incompatible implementation. > **We keep it as is for reproducibility in this branch** and will fix it in another PR for the main branch. --------- Co-authored-by: Guangming Sheng <shengguangming@bytedance.com> Co-authored-by: Guangming Sheng <petershengwhu@gmail.com>
chenjiaoAngel
added a commit
to chenjiaoAngel/verl
that referenced
this pull request
Nov 14, 2025
…roject#623) > [!WARNING] > As mentioned in verl-project#623 (comment), the implementation of gradient accumulation in verl has been only compatible with the sequence-mean loss, but all the DAPO experiments with the token-mean loss were run with the incompatible implementation. > **We keep it as is for reproducibility in this branch** and will fix it in another PR for the main branch. --------- Co-authored-by: Guangming Sheng <shengguangming@bytedance.com> Co-authored-by: Guangming Sheng <petershengwhu@gmail.com>
paolo328
added a commit
to paolo328/Verl
that referenced
this pull request
Nov 27, 2025
> [!WARNING] > As mentioned in verl-project/verl#623 (comment), the implementation of gradient accumulation in verl has been only compatible with the sequence-mean loss, but all the DAPO experiments with the token-mean loss were run with the incompatible implementation. > **We keep it as is for reproducibility in this branch** and will fix it in another PR for the main branch. --------- Co-authored-by: Guangming Sheng <shengguangming@bytedance.com> Co-authored-by: Guangming Sheng <petershengwhu@gmail.com>
TimurTaepov
pushed a commit
to giorgossideris/verl
that referenced
this pull request
Dec 20, 2025
…roject#623) > [!WARNING] > As mentioned in verl-project#623 (comment), the implementation of gradient accumulation in verl has been only compatible with the sequence-mean loss, but all the DAPO experiments with the token-mean loss were run with the incompatible implementation. > **We keep it as is for reproducibility in this branch** and will fix it in another PR for the main branch. --------- Co-authored-by: Guangming Sheng <shengguangming@bytedance.com> Co-authored-by: Guangming Sheng <petershengwhu@gmail.com>
vyomakesh0728
added a commit
to vyomakesh0728/verl
that referenced
this pull request
Jan 22, 2026
…roject#623) > [!WARNING] > As mentioned in verl-project#623 (comment), the implementation of gradient accumulation in verl has been only compatible with the sequence-mean loss, but all the DAPO experiments with the token-mean loss were run with the incompatible implementation. > **We keep it as is for reproducibility in this branch** and will fix it in another PR for the main branch. --------- Co-authored-by: Guangming Sheng <shengguangming@bytedance.com> Co-authored-by: Guangming Sheng <petershengwhu@gmail.com>
dreamyang-liu
pushed a commit
to dreamyang-liu/verl-sagemaker
that referenced
this pull request
Feb 21, 2026
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
6 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.
Warning
As mentioned in #623 (comment), the implementation of gradient accumulation in verl has been only compatible with the sequence-mean loss, but all the DAPO experiments with the token-mean loss were run with the incompatible implementation.
We keep it as is for reproducibility in this branch and will fix it in another PR for the main branch.