-
-
Notifications
You must be signed in to change notification settings - Fork 4.6k
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
[V1] Refactor model executable interface for all text-only language models #10374
base: main
Are you sure you want to change the base?
Conversation
Signed-off-by: Roger Wang <[email protected]>
👋 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:
🚀 |
Signed-off-by: Roger Wang <[email protected]>
Signed-off-by: Roger Wang <[email protected]>
Signed-off-by: Roger Wang <[email protected]>
Signed-off-by: Roger Wang <[email protected]>
Signed-off-by: Roger Wang <[email protected]>
Signed-off-by: Roger Wang <[email protected]>
Signed-off-by: Roger Wang <[email protected]>
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.
LGTM!
After you refactor the multi-modal models, can you update the base interface for vLLM models (class VllmModel
) to define get_input_embeddings
and also update the required arguments of forward
, then add a test to ensure that all of our models implement it?
@DarkLight1337 Thanks for the review! Yes eventually if we're going to enforce this interface, then I will update those model interfaces once the refactoring is done for all related models. I also want others to take a look at this PR before we proceed so I disabled auto-merge and CI for now. |
This PR refactors the interface of all text-only decoder language models for V1 VLM re-arch and
torch.compile
support. In particular, all model implementations on vLLM will need to meet the following requirements:get_input_embeddings(input_ids)
implemented inXYZModel
get_input_embeddings(input_ids)
implemented inXYZForCausalLM
,XYZForConditionalGeneration
, and/orXYZForClassification
inputs_embeds
is a required (with defaultNone
) parameter in theforward
function signature ofXYZModel
andXYZForCausalLM
,XYZForConditionalGeneration
, and/orXYZForClassification
This PR is a prerequisite of applying #9871 to all multimodal models on vLLM.