Closed antonbek89 closed 1 year ago
Hi, unfortunately not yet, I described all the problems here: https://github.com/exadel-inc/CompreFace/issues/519#issuecomment-848057021
Also, another problem would be that the raspberry pi's dont have AVX support, atleast not mine.
Any ARM device supports the AVX, but it's not a problem. If we take precompiled Tensorflow for ARM, it won't require AVX.
Any ARM device supports the AVX, but it's not a problem. If we take precompiled Tensorflow for ARM, it won't require AVX.
well this guy asked about the raspberry pi, which has a cpu without AVX support out of the box.
Raspberry pi has an ARM CPU, this is why I mentioned it. As soon as we build CompreFace for ARM, it should start working on all ARM devices, including Raspberry pi
@pospielov Any updates with ARM version?
Unfortunately no, I can't find a contributor for this, and I don't have time either
Unfortunately no, I can't find a contributor for this, and I don't have time either
What do you need?
I have a Jetson Nano 4gig I'd like to use this on. It was running DeepStack but even with CUDA support it does do near what the version of CompreFace I have installed in Home Assistant using just lots of CPU (40 core PowerEdge R620 running VMWare)
The biggest problem with running CompreFace on an ARM is that it uses libraries not compiled for ARM devices. For example, Tensorflow. When I last checked, Pillow wasn't supported as, well. But it looks like it supports now. So we need somebody who checks all the dependencies if they are supported by ARM. Then fix somehow all dependencies that don't support ARM. Make a new dockerfile (or update the old one) and finally, use Docker Buildx to set up CompreFace build for ARM devices.
The biggest problem with running CompreFace on an ARM is that it uses libraries not compiled for ARM devices.
For example, Tensorflow. When I last checked, Pillow wasn't supported as, well. But it looks like it supports now.
So we need somebody who checks all the dependencies if they are supported by ARM.
Then fix somehow all dependencies that don't support ARM.
Make a new dockerfile (or update the old one)
and finally, use Docker Buildx to set up CompreFace build for ARM devices.
I have all the hardware. But don't know enough programming to do all of that. Although I'm no slouch at the cli and I'm not afraid to experiment. You'd have to walk me through it step by step.
@LordNex Could you test this build on your ARM device? https://github.com/exadel-inc/CompreFace/pull/1018
@LordNex Could you test this build on your ARM device?
Yea it'll take me a bit. I've got to take the case apart to get the SD card out so I can load it back up. But yea I'll get it tested as soon as I can.
What are the performance? Raspberry Pi 4 4gb is enough?
I’m currently running Compreface on a system with an AArch64 (ARM 64-bit) architecture and encountered compatibility issues with some services. Specifically, I am having trouble running the compreface-core service on this architecture.
I’ve successfully managed to run other containers on my AArch64 setup by using multi-platform support via docker buildx, but I am unsure if compreface-core is compatible or if there are any known issues or requirements for running it on ARM-based systems.
Could you please provide information on the following:
AArch64 Support: Does Compreface officially support ARM 64-bit architecture? Are there specific steps or configurations required to run compreface-core on an AArch64 system?
Docker Images: Are there pre-built Docker images for ARM 64-bit architecture for Compreface services? If not, are there recommended practices or solutions for building these images from source?
Known Issues: Are there any known issues or limitations when running Compreface services on ARM-based systems?
Documentation: Is there any additional documentation or community resources for running Compreface on non-x86 architectures?
Any guidance or assistance would be greatly appreciated as it will help in ensuring compatibility and smooth operation of Compreface on ARM-based systems.
Thank you for your time and support!
The linux version Linux raspberrypi 5.10.60-v8+ #1449 SMP PREEMPT Wed Aug 25 15:01:33 BST 2021 aarch64 GNU/Linux