vllm-project
[BugFix] Propagate 'trust_remote_code' setting in internvl and minicpmv
It's a bit tricky for MiniCPM-V since ModelConfig is not passed into MiniCPMV but used in _get_image_bounds().
PR Checklist (Click to Expand)
Thank you for your contribution to vLLM! Before submitting the pull request, please ensure the PR meets the following criteria. This helps vLLM maintain the code quality and improve the efficiency of the review process.
PR Title and Classification
Only specific types of PRs will be reviewed. The PR title is prefixed appropriately to indicate the type of change. Please use one of the following:
[Bugfix]for bug fixes.[CI/Build]for build or continuous integration improvements.[Doc]for documentation fixes and improvements.[Model]for adding a new model or improving an existing model. Model name should appear in the title.[Frontend]For changes on the vLLM frontend (e.g., OpenAI API server,LLMclass, etc.)[Kernel]for changes affecting CUDA kernels or other compute kernels.[Core]for changes in the core vLLM logic (e.g.,LLMEngine,AsyncLLMEngine,Scheduler, etc.)[Hardware][Vendor]for hardware-specific changes. Vendor name should appear in the prefix (e.g.,[Hardware][AMD]).[Misc]for PRs that do not fit the above categories. Please use this sparingly.
Note: If the PR spans more than one category, please include all relevant prefixes.
Code Quality
The PR need to meet the following code quality standards:
- We adhere to Google Python style guide and Google C++ style guide.
- Pass all linter checks. Please use
format.shto format your code. - The code need to be well-documented to ensure future contributors can easily understand the code.
- Include sufficient tests to ensure the project to stay correct and robust. This includes both unit tests and integration tests.
- Please add documentation to
docs/source/if the PR modifies the user-facing behaviors of vLLM. It helps vLLM user understand and utilize the new features or changes.
Notes for Large Changes
Please keep the changes as concise as possible. For major architectural changes (>500 LOC excluding kernel/data/config/test), we would expect a GitHub issue (RFC) discussing the technical design and justification. Otherwise, we will tag it with rfc-required and might not go through the PR.
What to Expect for the Reviews
The goal of the vLLM team is to be a transparent reviewing machine. We would like to make the review process transparent and efficient and make sure no contributor feel confused or frustrated. However, the vLLM team is small, so we need to prioritize some PRs over others. Here is what you can expect from the review process:
- After the PR is submitted, the PR will be assigned to a reviewer. Every reviewer will pick up the PRs based on their expertise and availability.
- After the PR is assigned, the reviewer will provide status update every 2-3 days. If the PR is not reviewed within 7 days, please feel free to ping the reviewer or the vLLM team.
- After the review, the reviewer will put an
action-requiredlabel on the PR if there are changes required. The contributor should address the comments and ping the reviewer to re-review the PR. - Please respond to all comments within a reasonable time frame. If a comment isn't clear or you disagree with a suggestion, feel free to ask for clarification or discuss the suggestion.
Thank You
Finally, thank you for taking the time to read these guidelines and for your interest in contributing to vLLM. Your contributions make vLLM a great tool for everyone!
👋 Hi! Thank you for contributing to the vLLM project.
Just a reminder: PRs would not trigger full CI run by default. Instead, it would only run fastcheck CI which starts running only a small and essential subset of CI tests to quickly catch errors. You can run other CI tests on top of those by going to your fastcheck build on Buildkite UI (linked in the PR checks section) and unblock them. If you do not have permission to unblock, ping simon-mo or khluu to add you in our Buildkite org.
Once the PR is approved and ready to go, your PR reviewer(s) can run CI to test the changes comprehensively before merging.
To run CI, PR reviewers can do one of these:
- Add
readylabel to the PR - Enable auto-merge.
🚀
![github-actions[bot] avatar](https://avatars.githubusercontent.com/in/15368?v=4 "Published by a pub.dev verified publisher"){kind=small}
Sorry for the delay. I still need to get the tests passed. I'll let you know once it's working.
@DarkLight1337 PTAL. I have to register another plugin for image_bounds aux data.
Rather than using another plugin, I suggest that you override the default input mapper (see Fuyu model for an example).
Rather than using another plugin, I suggest that you override the default input mapper (see Fuyu model for an example).
I tried that idea, but input mapper can't access token ids. Am I missing anything?
I mean that you can override the input mapper so that it accepts the image bounds from the input processor without having to define another plugin to handle it.
I mean that you can override the input mapper so that it accepts the image bounds from the input processor without having to define another plugin to handle it.
map_input is tied to plugins:
https://github.com/vllm-project/vllm/blob/3118f63385c0d767fba8b6d2039fc35440678da9/vllm/multimodal/registry.py#L116
I would get a Unknown multi-modal data type: image_bounds error if i didn't register a plugin.
Oh, I forgot about that part... perhaps you can change the type of data passed to "image" key instead. For example,
multi_modal_data["image"] = {"data": image, "bounds": image_bounds}
and override the input mapper to handle this case.
multi_modal_data["image"] = {"data": image, "bounds": image_bounds}
I tried it out and it's also very awkward:
multi_modal_data["image"] = {"data": List[Image], "bounds": image_bounds} won't work with --limit-mm-per-prompt, since the image input is no longer a list.
On the other hand, multi_modal_data["image"] = [{"data": List[Image], "bound": image_bound}] also won't work, since len(image_bounds) is not always the same as len(images). image_bounds is computed completely on prompt tokens.
On the other hand,
multi_modal_data["image"] = [{"data": List[Image], "bound": image_bound}]also won't work, sincelen(image_bounds)is not always the same aslen(images).image_boundsis computed completely on prompt tokens.
If the number of images represented by the prompt isn't the same as the number of input images, it is probably a user error. IMO, an error should be thrown in that case anyway.
On the other hand,
multi_modal_data["image"] = [{"data": List[Image], "bound": image_bound}]also won't work, sincelen(image_bounds)is not always the same aslen(images).image_boundsis computed completely on prompt tokens.If the number of images represented by the prompt isn't the same as the number of input images, it is probably a user error. IMO, an error should be thrown in that case anyway.
It's related to the "slice" concept used in the model which I am not familiar with https://github.com/vllm-project/vllm/blob/a91165f7afe08ad47750c6d5270471aa0242e27f/vllm/model_executor/models/minicpmv.py#L298. The length of image_bounds can be much larger than number of images.
How's the latest version with input mapper? "image_bounds" is carried around with every input image which is less than ideal. But otherwise pretty clean to me.