Skip to content
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

Which version of InternVL does TensorRT-llm 1.5 support ? #2578

Open
spacegrass opened this issue Dec 16, 2024 · 3 comments
Open

Which version of InternVL does TensorRT-llm 1.5 support ? #2578

spacegrass opened this issue Dec 16, 2024 · 3 comments
Assignees
Labels
Investigating LLM API/Workflow triaged Issue has been triaged by maintainers

Comments

@spacegrass
Copy link

From TensorRT-LLM code
Image

The image preprocess is internvl 1.2 version.
Form InternVL 1.5/2.0 the image preprocess like this:
Image

So, does TensorRT-llm 1.5(the lastest version) support InternVL 2.0 model ?

@nv-guomingz
Copy link
Collaborator

Hi @spacegrass the latest code already declared the internvl2 supporting. Please refer to

https://github.com/NVIDIA/TensorRT-LLM/tree/main/examples/multimodal#internvl2

@nv-guomingz nv-guomingz added the triaged Issue has been triaged by maintainers label Dec 18, 2024
@spacegrass
Copy link
Author

Hi @spacegrass the latest code already declared the internvl2 supporting. Please refer to

https://github.com/NVIDIA/TensorRT-LLM/tree/main/examples/multimodal#internvl2

--- We have use the script to convert hf-model 2 trt-llm model, I find the image process code in the tensorrt-llm main & 0.15.0 branch is the InternVL 1.2 version. the processed image shape is [1, 3, 448, 448], bute the InternVL 2.0 add dynamic split so the default shape is [3, 3, 448, 448]

@nv-guomingz
Copy link
Collaborator

@sunnyqgg could u please double confirm this issue?

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
Investigating LLM API/Workflow triaged Issue has been triaged by maintainers
Projects
None yet
Development

No branches or pull requests

3 participants