slashgob Posted April 3 Share Posted April 3 14 hours ago, yayitazale said: You are missing the usb passtrough. I suggest you to start again with a clean template without deleting this: Also, I see in the config file that you are trying to use a yolo model with the coral and that is not posible. As you can read in the docs, the best option is to use the default model that is just fine, so you should comment or delete this part: model: path: /config/model_cache/tensorrt/yolov7-320.trt input_tensor: nchw input_pixel_format: rgb width: 320 height: 320 Yolo models are great for nvidia or openvino detectors, but the coral is a different chip that has its own requirements to create a model. If you want to use another custom modelwith the coral, it must be a tflite model. You can find here most of the pretrained models. It was the model I had configured for when I was using the GPU. I commented this part out and its working! Thanks very much for your help 1 1 Quote Link to comment
awediohead Posted April 3 Share Posted April 3 --gpus=all --shm-size=256mb --mount type=tmpfs,target=/tmp/cache,tmpfs-size=100000000 --restart unless-stopped Can someone explain what the --gpus=all part does of the above Extra Parameters? When is it necessary / useful as its not in the default ? Quote Link to comment
neleh Posted April 4 Share Posted April 4 18 hours ago, yayitazale said: Frigate template is pointing to the "latest" tag so yo you should check if you have change that. Any way, you should reads the changelog and the updated docs to find out what changes you need to make to your settings has there are many new things. Also it is recommended to ensure that you are using all the required new variables in the template or reinstall it from scratch deleting you actual container. Thanks, you're right. It turns out my Community Apps didn't update, which then caused the Docker page to show everything uptodate when they are not. Quote Link to comment
yayitazale Posted April 4 Author Share Posted April 4 (edited) 12 hours ago, awediohead said: --gpus=all --shm-size=256mb --mount type=tmpfs,target=/tmp/cache,tmpfs-size=100000000 --restart unless-stopped Can someone explain what the --gpus=all part does of the above Extra Parameters? When is it necessary / useful as its not in the default ? That parameter is the proposed one to launch frigate with docker-run with a nvidia gpu pass-trough: https://docs.frigate.video/configuration/hardware_acceleration#docker-run-cli---nvidia-gpu As we are on Unraid, there is a more "standar" way to pass-trough defined on the nvidia driver plugin, and it is the one implemented on the template: Edited April 4 by yayitazale 1 Quote Link to comment
awediohead Posted April 4 Share Posted April 4 Thanks for clarifying and confirming it's not something I need for now. It had been suggested to me as necessary previously, hence my question to double check. Cheers 1 Quote Link to comment
Zeze21 Posted Sunday at 08:41 PM Share Posted Sunday at 08:41 PM On 4/4/2024 at 1:59 PM, yayitazale said: That parameter is the proposed one to launch frigate with docker-run with a nvidia gpu pass-trough: https://docs.frigate.video/configuration/hardware_acceleration#docker-run-cli---nvidia-gpu As we are on Unraid, there is a more "standar" way to pass-trough defined on the nvidia driver plugin, and it is the one implemented on the template: Actually I have a question regarding this as frigate in my case seems to "grab" all GPUs available and this is what i do NOT want. I have some dedicated GPUs for VMs and if frigate uses those GPUs the VMs won't start, therefore I want to limit frigate to just one GPU. But somehow this seems impossible to me: here's my Frigate Docker Config: docker run -d --name='frigate' --net='host' --privileged=true -e TZ="Europe/Berlin" -e HOST_OS="Unraid" -e HOST_HOSTNAME="Server" -e HOST_CONTAINERNAME="frigate" -e 'TCP_PORT_5000'='5000' -e 'TCP_PORT_8554'='8554' -e 'FRIGATE_RTSP_PASSWORD'='***' -e 'NVIDIA_VISIBLE_DEVICES'='GPU-bd161e6e-9b15-d71a-ec49-8c085ea02ad1' -e 'NVIDIA_DRIVER_CAPABILITIES'='compute,utility,video' -e 'TCP_PORT_8555'='8555' -e 'UDP_PORT_8555'='8555' -e 'TCP_PORT_1984'='1984' -l net.unraid.docker.managed=dockerman -l net.unraid.docker.webui='http://[IP]:[PORT:5000]' -l net.unraid.docker.icon='https://raw.githubusercontent.com/yayitazale/unraid-templates/main/frigate.png' -v '/mnt/user/appdata/frigate':'/config':'rw' -v '/mnt/user/Media/frigate':'/media/frigate':'rw' -v '/etc/localtime':'/etc/localtime':'rw' --device='/dev/bus/usb/006/002' --runtime=nvidia --shm-size=256mb --mount type=tmpfs,target=/tmp/cache,tmpfs-size=1000000000 --restart unless-stopped 'ghcr.io/blakeblackshear/frigate:stable' 440515ff723a15052068c74656fef26f189832b762fe93eb1766155f8130e599 The command finished successfully! The only way is to start the VMs before I start frigate - this is not an option for me. With this config and one VM started before frigate, frigate still shows up 2 GPUs: Please how can i remove the rtx 3060? Or in other words how can i limit frigate to the 1030? The 1030 has the GPU id mentioned above in the frigate docker config. Thank you for your advice! 1 Quote Link to comment
BrandonG777 Posted 6 minutes ago Share Posted 6 minutes ago On 4/3/2024 at 4:12 AM, slashgob said: It was the model I had configured for when I was using the GPU. I commented this part out and its working! Thanks very much for your help Good looking out! I bet this was my issue all along. Quote Link to comment
Recommended Posts
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.