-
-
Notifications
You must be signed in to change notification settings - Fork 6.3k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
[Model] Add Support for Multimodal Granite Models #10291
Conversation
👋 Hi! Thank you for contributing to the vLLM project. 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:
🚀 |
b31ef20
to
3e1995e
Compare
This looks a bit overcomplicated, and can be easily confused with |
The reason why we set |
d9592fe
to
c77bd1f
Compare
Sounds good, thanks @DarkLight1337! I moved most of the vision feature layer handling out to inference time, added the flag for getting all of the hidden states + selecting from them after, and fixed the initialization behavior so that we only load up to the deepest layer needed if there are multiple feature layers also |
vllm/model_executor/models/clip.py
Outdated
self.encoder = CLIPEncoder( | ||
config=config, | ||
quant_config=quant_config, | ||
num_hidden_layers_override=num_hidden_layers_override, | ||
prefix=f"{prefix}.encoder", |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Seems like an unnecessary formatting change.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Yup! I put it back the way it was, accidentally deleted some trailing commas + auto formatted in some places 😄
44e5233
to
20d6de5
Compare
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Otherwise LGTM
cc @ywang96 in case you're planning to refactor related code in the vision encoders. |
Signed-off-by: Alex-Brooks <Alex.Brooks@ibm.com>
Signed-off-by: Alex-Brooks <Alex.Brooks@ibm.com>
Signed-off-by: Alex-Brooks <Alex.Brooks@ibm.com> Fix post norm handline for multi feature layers Signed-off-by: Alex-Brooks <Alex.Brooks@ibm.com>
Signed-off-by: Alex-Brooks <Alex.Brooks@ibm.com>
Signed-off-by: Alex-Brooks <Alex.Brooks@ibm.com>
Signed-off-by: Alex-Brooks <Alex.Brooks@ibm.com>
Co-authored-by: Cyrus Leung <cyrus.tl.leung@gmail.com> Signed-off-by: Alex-Brooks <Alex.Brooks@ibm.com>
Co-authored-by: Cyrus Leung <cyrus.tl.leung@gmail.com> Signed-off-by: Alex-Brooks <Alex.Brooks@ibm.com>
Signed-off-by: Alex-Brooks <Alex.Brooks@ibm.com>
Signed-off-by: Alex-Brooks <Alex.Brooks@ibm.com>
Head branch was pushed to by a user without write access
593aa66
to
31d98de
Compare
Signed-off-by: Alex-Brooks <Alex.Brooks@ibm.com> Co-authored-by: Cyrus Leung <cyrus.tl.leung@gmail.com> Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com>
Signed-off-by: Alex-Brooks <Alex.Brooks@ibm.com> Co-authored-by: Cyrus Leung <cyrus.tl.leung@gmail.com> Signed-off-by: Maxime Fournioux <55544262+mfournioux@users.noreply.github.com>
Signed-off-by: Alex-Brooks <Alex.Brooks@ibm.com> Co-authored-by: Cyrus Leung <cyrus.tl.leung@gmail.com>
Signed-off-by: Alex-Brooks <Alex.Brooks@ibm.com> Co-authored-by: Cyrus Leung <cyrus.tl.leung@gmail.com>
This PR adds the main architectural change needed to support upcoming multimodal granite models from IBM, which are a variant of llava-next. To this end, we need to allow passing a list of
feature_sample_layers
, which are indices in the visual encoder whose hidden states will be concatenated and used as the visual encoder output as an alternative to the last layer. This change should not affect any existing llava next models, and is made in all of the supported visual encoders for Llava-based models in vLLM (clip/siglip/pixtral) to keep things as generic as possible.These models have not been released quite yet, and we have not yet ported the change to transformers, which is why there are no new tests added. However, I have run my own benchmarks to ensure the results in vLLM are identical in quality to our own implementation, and verified that the
feature_sample_layers
can be passed through each supported visual encoder type as a sanity check. Once the models are out and supported intransformers
, I can add it to the corresponding test here!FYI @njhill
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,LLM
class, 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:
format.sh
to format your code.docs/source/
if the PR modifies the user-facing behaviors of vLLM. It helps vLLM user understand and utilize the new features or changes.Adding or changing kernels
Each custom kernel needs a schema and one or more implementations to be registered with PyTorch.
Tensors
require meta-functions. Meta-functions should be implemented and registered in python so that dynamic dims can be handled automatically. See above documents for a description of meta-functions.torch.libary.opcheck()
to test the function registration and meta-function for any registered ops. Seetests/kernels
for examples.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:
action-required
label on the PR if there are changes required. The contributor should address the comments and ping the reviewer to re-review the PR.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!