verl-project/verl

Fused CE loss integration

Open

#97 opened on Jan 12, 2025

View on GitHub
 (2 comments) (1 reaction) (0 assignees)Python (21,533 stars) (3,940 forks)auto 404
call for contributionhelp wanted

Description

Integrate it with main stream models: https://github.com/apple/ml-cross-entropy so that model with large vocab size uses much less memory

Contributor guide

Tech stack
pythonpytorch
Domain
machine learningperformance
Issue type
feature
DifficultyEstimated implementation difficulty for a new contributor, from 1 for very small changes to 5 for expert-level work.
3
Estimated timeA rough time range for an experienced contributor to investigate, implement, test, and prepare a pull request.
half day
Activity statusHow available the issue appears right now: fresh, active, stale, blocked, or waiting on maintainer input.
fresh
ClarityHow clearly the issue explains the expected change, acceptance criteria, and next step.
needs investigation
Prerequisites
PyTorchCUDAcross entropy loss
Newbie friendlinessA 1-100 score estimating how approachable this issue is for first-time contributors.
40
Research direction
First, study the apple/ml cross entropy repository to understand the fused CE loss kernel and its API. Then, explore verl's model code, likely in the 'verl' directory, to identify where standard cross entropy loss is computed for large vocabulary models. Look for issues or discussions in the comments to see if maintainers specified which models to target. Consider opening a discussion to clarify the scope before implementation.