I’ve been following the “set up inference server” roboflow tutorial (X9jt8qb_igo) to the letter, trying to get local inferencing from a roboflow model working on the jetson nano dev kit.
I setup things OK on the jetson nano 4GB using JetPack 4.6.1, and have the local inference server running on the jetson with docker with the following command:
docker run --net=host --gpus all roboflow/inference-server:jetson
The server appears to run OK. In another terminal, I’m trying to run a test inference on a local image using the following command (from the tutorial):
base64 1689278366.jpg | curl -d @- “http://127.0.0.1:9001/jsi/8?api_key=VLOziA70pvIDNWDTQ66H”
Initially, on the device I see that it can load the model OK, but then I get an impassable error.
Can someone please assist with this issue or attempt to reproduce it? As-is, the getting started tutorial does not work out of the box as a result with the jetson nano.
Might be connected to my issue?
Hey, @Charles_Modrich, sorry to hear of the troubles. Which tutorial are you following? Is it this YouTube video?
Charlie’s team member here. Yes, that video and blog post / how-to.
Since the image of the error doesn’t seem to be working on the original post:
Hi Joesph! Any thoughts on this one? Happy to test with your team to update the documentation for others.
@NickJSI As mentioned above, your issue might be the same as ours. You got the same error message we got when following the written docs on a Jetson Nano trying to deploy Roboflow 3.0 models.
According to the answer of @Paul (Error deploying custom trained model on NVIDIA Jetson Xavier - #15 by Paul) you could follow the new documentation (NVIDIA Jetson - Roboflow Docs), but I have bumped into quota issues that might be also the case for you and will require someone from roboflow to solve.
Wanted to update this thread that there are new docs and new Jetson images that should resolve this issue.
New Docker Images:
This topic was automatically closed 10 days after the last reply. New replies are no longer allowed.