-
-
Notifications
You must be signed in to change notification settings - Fork 5.9k
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] Clean up MiniCPMV #10751
[Model] Clean up MiniCPMV #10751
Conversation
Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
👋 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:
🚀 |
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.
Overall LGTM! Just have an idea about further cleanup.
8e8b394
to
39b8fc3
Compare
Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
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 and thanks so much for getting rid of LLMWrapper
!
tests/models/decoder_only/vision_language/vlm_utils/model_utils.py
Outdated
Show resolved
Hide resolved
Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
Somehow this breaks MiniCPMV 2.0, not sure why. |
Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
I fixed the problem - turns out scaling was incorrectly applied twice. |
Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
@DarkLight1337 looks like it's breaking model loading
|
Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk> Signed-off-by: Andrew Feldman <afeldman@neuralmagic.com>
Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
This PR cleans up the code of MiniCPMV, simplifying the weight loading code and removing the need to use
LLMWrapper
.