Hi there ​@Drouen! Welcome on board!

I’m wondering based on your other replies, if you’re using Windows? Or a non-Linux operating system, anyway. What’s your OS, and its version?

Windows 11 i skipped WSL and have a side partition now with ubuntu 22. Everything installs no problems only thing is now. I got the m.2 on a blazing port but my pc cannot see the m.2. i am sure that the voltage is enough so, how can i install the driver correctly?

 

After tweaking a bit the blazing port does the trick!.

so for fokes out there having trouble that they do not see the lspci look at power input. for the metis m.2 the port to hyper m.2 is a no go, blazing port does the trick.

Ah, excellent work ​@Drouen! What kind of tweaks did the blazing port need, out of interest?

 

Check out that sweet kit!

What project are you building with it?

I want to make it ‘captain of the ship’ so that it controls my gpu and cpu. Can that be done?

what am i missing in this command?

./inference.py rtsp://<user>:<password>@192.168.0.100:8554/1

 

his answer is:

usage: deploy.py -h] --build-root PATH] --data-root PATH]
                 Â--model MODEL | --models-only | --pipeline-only]
                 Â--mode {QUANTIZE,QUANTIZE_DEBUG,QUANTCOMPILE,PREQUANTIZED}]
                 U--metis {auto,none,pcie,m2}]
                 m--pipe {gst,torch,torch-aipu}] Â--export]
                 [--loglevel LEVEL] t--logfile PATH] x--logtimestamp]
                  --brief-logging] L-q | -v]
                 m--enable-vaapi | --disable-vaapi | --auto-vaapi]
                 v--enable-opencl | --disable-opencl | --auto-opencl]
                 |--enable-opengl | --disable-opengl | --auto-opengl]
                 d--aipu-cores {1,2,3,4}]
                 Â--num-cal-images NUM_CAL_IMAGES]
                 network
deploy.py: error: "Invalid network

i tried diffrent paths and also without

Is it the syntax that's the problem? You've got the source (your RTSP feed) but not the model.

 

So along the lines of: 

./inference.py <network-name> <input-source>

Ok, can you give me an example, because there are loads of maps in voyager-sdk .

For the record its a cctv so face recognition and the whole shabam can go over it, if you have the perfect model.

 

Best regards

Ha ha! Yeah, there’s a lot going on and a lot of options!

Does this section help? It’s got some simple examples, with the models and the source (the latter being your RTSP feed): https://github.com/axelera-ai-hub/voyager-sdk/blob/release/v1.2.5/docs/tutorials/quick_start_guide.md#run-a-metis-accelerated-pipeline

(By the way, I like this project! It’s something I’d like to set up at home, as part of my automation system!)

Top tip!