Roboflow Inference, ModelDependencyMissing: CLIP

🤝 Community Help
1

Hello, I am trying to get Roboflow inference clip to run on my Jetson Orin AGX with Jetpack 6.2.1.
At first I was not aware that we use 6.2.1 instead of 6.2.0, so I went with it.
Building the server worked and I can run example files which do not require additionally installed models such as clip, florence2 etc.
Now when I run the server the output says:
ModelDependencyMissing: Your inference configuration does not support CLIP model. Use pip install ‘inference[clip]’ to install missing requirements
I have installed it in the docker, but it still does not change.
I cannot seem to find anything related to the ModelDependencyMissing issue.
Is it the Jetpack version that is screwing everything up or is there something I can try to do to make it work?

Commands used to build:

docker build -f docker/dockerfiles/Dockerfile.onnx.jetson.6.2.0\ -t roboflow/roboflow-inference-server-trt-jetson-6.2.0 .

sudo docker run --privileged --net=host --runtime=nvidia --mount source=roboflow,target=/tmp/cache -e NUM_WORKERS=1 roboflow/roboflow-inference-server-trt-jetson-6.2.0:latest

Thank you in advance!

2

Hi @Jakob ,

You have done the right thing by installing clip dependencies in the docker, in order to enable it you also need to set the environmental variable, add -e CORE_MODEL_CLIP_ENABLED=True to your docker run params

Hope this helps!

Grzegorz

3

This topic was automatically closed 21 days after the last reply. New replies are no longer allowed.