SP on VLMs #71
Merged
SP on VLMs #71
Conversation
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
added 3 commits
July 30, 2025 16:35
- Remove data printing functions from SFT and DPO trainers for better performance - Replace 360-example-vl.sh with separate SFT and DPO training scripts - Add SFT visual-language demo dataset (data/sft-vl-demo/) - Update dataset configuration to support new data structure
… code style - Add multimodal_forwards module to centrally manage multimodal model forward logic - Extract and optimize forward function implementations for Qwen2 VL and Qwen2.5 VL - Improve sequence_parallel related code structure
This was referenced Sep 28, 2025
Closed
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 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Outdated suggestions cannot be applied.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.
What does this PR do?
Extends existing text sequence parallelism to support multimodal (image + text) training for SFT and DPO with padding removal optimization.
Key Changes
Multimodal sequence parallelism for SFT/DPO: Added support for vision-language models in both Supervised Fine-Tuning (SFT) and Direct Preference Optimization (DPO) training while maintaining backward compatibility with text-only training
Remove padding optimization: Implemented dynamic padding removal to reduce memory usage and improve training efficiency
Enhanced parallel communication: Updated tensor distribution patterns for multimodal data