[Optimization] Refine row parallel bias and nranks and moe all_reduce #5247
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
|
Thanks for your contribution! |
Contributor
There was a problem hiding this comment.
Pull request overview
This PR refactors tensor parallelism-related code by standardizing variable naming and improving bias handling for distributed training. The changes focus on code consistency and correctness without altering the core functionality.
- Standardizes variable naming from
nrankstotp_sizeacross multiple modules for better clarity - Introduces special handling for row-parallel bias division in tensor parallelism
- Removes unused variables and simplifies conditional logic in RowParallelLinear
Reviewed changes
Copilot reviewed 9 out of 9 changed files in this pull request and generated 4 comments.
Show a summary per file
| File | Description |
|---|---|
| fastdeploy/model_executor/utils.py | Adds tp_row_bias attribute handling to divide bias by tensor parallel size during weight loading |
| fastdeploy/model_executor/models/qwen3moe.py | Removes unused self.nranks variable |
| fastdeploy/model_executor/models/qwen3.py | Renames nranks to tp_size for consistency |
| fastdeploy/model_executor/models/qwen2.py | Removes unused self.nranks variable |
| fastdeploy/model_executor/models/ernie4_5_moe.py | Removes unused self.nranks variable |
| fastdeploy/model_executor/layers/mtp_linear.py | Renames self.nranks to self.tp_size |
| fastdeploy/model_executor/layers/lm_head.py | Renames self.nranks to self.tp_size |
| fastdeploy/model_executor/layers/linear.py | Renames variables, removes unused field, adds bias attribute handling, and simplifies logic |
| fastdeploy/model_executor/layers/backends/intel_hpu/attention/hpu_attn_backend.py | Renames self.nranks to self.tp_size |
Codecov Report❌ Patch coverage is Additional details and impacted files@@ Coverage Diff @@
## develop #5247 +/- ##
==========================================
Coverage ? 59.92%
==========================================
Files ? 317
Lines ? 38774
Branches ? 5843
==========================================
Hits ? 23234
Misses ? 13703
Partials ? 1837
Flags with carried forward coverage won't be shown. Click here to find out more. ☔ View full report in Codecov by Sentry. 🚀 New features to boost your workflow:
|
yuanlehome
changed the title
| @@ -211,7 +211,7 @@ def __init__( | |||
| self.speculate_max_draft_token_num: int = llm_config.speculative_config.num_speculative_tokens | |||
| self.keep_pd_step_flag: bool = llm_config.speculative_config.model_type == "mtp" | |||
| self.rank: int = llm_config.parallel_config.tensor_parallel_rank | |||
Collaborator
There was a problem hiding this comment.
self.rank 的命名也不准确，应该直接叫self.tp_rank
1 task
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 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.
Motivation
Modifications
Usage or Command
无。
Accuracy Tests
无。
Checklist
[FDConfig],[APIServer],[Engine],[Scheduler],[PD Disaggregation],[Executor],[Graph Optimization],[Speculative Decoding],[RL],[Models],[Quantization],[Loader],[OP],[KVCache],[DataProcessor],[BugFix],[Docs],[CI],[Optimization],[Feature],[Benchmark],[Others],[XPU],[HPU],[GCU],[DCU],[Iluvatar],[Metax]]pre-commitbefore commit.releasebranch, make sure the PR has been submitted to thedevelopbranch, then cherry-pick it to thereleasebranch with the[Cherry-Pick]PR tag.