[vl]remove duplicated load logic #2744
Merged
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! |
|
|
a6690a6 to
05b4eda
Compare
yuanlehome
reviewed
Jul 8, 2025
yuanlehome
reviewed
Jul 8, 2025
70afc2c to
c64018b
Compare
There was a problem hiding this comment.
Pull Request Overview
This PR removes duplicated distributed environment initialization logic from the vision-language model runner by reusing the centralized loading logic from worker_process.py. The main purpose is to consolidate configuration initialization and reduce code duplication across the codebase.
Key changes include:
- Extracted common parallel configuration initialization into reusable functions
- Replaced custom model building logic in VL model runner with centralized configuration initialization
- Updated function signatures to accept rank parameters for proper distributed setup
Reviewed Changes
Copilot reviewed 11 out of 11 changed files in this pull request and generated 6 comments.
Show a summary per file
| File | Description |
|---|---|
| fastdeploy/worker/worker_process.py | Extracted distributed initialization and parallel config setup into reusable functions, updated function signatures |
| fastdeploy/worker/vl_gpu_model_runner.py | Removed duplicated model building logic and reused centralized initialization from worker_process.py |
| fastdeploy/rl/rollout_config.py | Updated function call to pass required rank parameters |
| fastdeploy/model_executor/models/utils.py | Added new expert ID placeholders for text and image experts |
| fastdeploy/model_executor/models/tp_utils.py | Enhanced key expansion logic to handle new expert types and improved error handling |
| fastdeploy/model_executor/models/ernie4_5_vl/modeling_resampler.py | Added prefix name support and custom state dict loading |
| fastdeploy/model_executor/models/ernie4_5_vl/ernie4_5_vl_moe.py | Restructured model initialization and added tensor parallel mappings |
| fastdeploy/model_executor/models/ernie4_5_vl/dfnrope/modeling.py | Added prefix name support and custom state dict loading |
| fastdeploy/model_executor/models/ernie4_5_moe.py | Updated tensor parallel mapping function calls |
| fastdeploy/model_executor/model_loader.py | Added VL model support to model registry |
| fastdeploy/config.py | Updated type annotations for MoE configuration fields |
Comments suppressed due to low confidence (1)
fastdeploy/worker/vl_gpu_model_runner.py:159
- Missing space after comma in parameter list - should be '_init_image_preprocess(self, vision_config)'
def _init_image_preprocess(self,vision_config) -> None:
fastdeploy/model_executor/models/ernie4_5_vl/ernie4_5_vl_moe.py
Outdated
Show resolved
Hide resolved
fastdeploy/model_executor/models/ernie4_5_vl/ernie4_5_vl_moe.py
Outdated
Show resolved
Hide resolved
fastdeploy/model_executor/models/ernie4_5_vl/ernie4_5_vl_moe.py
Outdated
Show resolved
Hide resolved
yuanlehome
reviewed
Jul 11, 2025
yuanlehome
reviewed
Jul 11, 2025
yuanlehome
reviewed
Jul 11, 2025
d95f493 to
b7c1c5f
Compare
yuanlehome
approved these changes
Jul 12, 2025
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.
Reuse the loading logic from fastdeploy/worker/worker_process.py