Merged
Conversation
Contributor
|
Warning You have reached your daily quota limit. Please wait up to 24 hours and I will start processing your requests again! |
Collaborator
Author
|
/unittest-all |
Collaborator
Author
|
/gemini review |
Summary
Skipped
Tests
Github Test Reporter by CTRF 💚 |
There was a problem hiding this comment.
Pull request overview
Adds support and compatibility patches for the Qwen3.5 model family, including updates to trainer data preparation, model monkey-patching, vLLM integration, and Megatron distributed checkpoint metadata handling.
Changes:
- Add Qwen3.5-specific monkey patches (forward overrides, ulysses sequence-parallel hooks, FLOPs estimator wiring).
- Adjust verl trainer preprocessing to use an “empty” HF model for multimodal rope-index computation.
- Extend Megatron distributed checkpoint save/load metadata support and widen several dependency constraints.
Reviewed changes
Copilot reviewed 11 out of 11 changed files in this pull request and generated 4 comments.
Show a summary per file
| File | Description |
|---|---|
trinity/trainer/verl/verl_trainer.py |
Builds an empty HF model (no weights) and passes it into to_data_proto for rope-index handling. |
trinity/trainer/verl/verl_config.py |
Adds Megatron distributed-checkpoint optimizer reshaping toggles. |
trinity/trainer/verl/utils.py |
Switches multimodal rope-index path from processor-based to model-based and removes local hf_processor shim. |
trinity/trainer/verl/monkey_patch.py |
Refactors fused-backend forward patching and adds Qwen3.5 model-type handling & related patches. |
trinity/trainer/verl/megatron_checkpoint_manager.py |
Adds content metadata generation/plumbing for dist-checkpointing save/load (compat workaround pending verl upgrade). |
trinity/common/patch/qwen3_5.py |
Introduces Qwen3.5 forward implementations for torch/triton backends plus ulysses linear-attn decorator. |
trinity/common/models/vllm_patch/worker_patch.py |
Expands supported vLLM version range for the prompt-logprobs patch. |
trinity/common/models/vllm_patch/__init__.py |
Adds a transformers>=5 compatibility monkey-patch for PreTrainedConfig.__init__. |
trinity/common/models/vllm_model.py |
Adjusts prompt truncation plumbing across vLLM versions (SamplingParams vs tokenization_kwargs). |
pyproject.toml |
Relaxes megatron-core / transformer-engine / flash-attn version constraints. |
.github/workflows/docker/docker-compose.yaml |
Updates CI docker image tags used by the workflow. |
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
You can also share your feedback on Copilot code review. Take the survey.
pan-x-c
approved these changes
Mar 11, 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
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
Checklist
Please check the following items before code is ready to be reviewed.