TDA Posted October 31, 2023 Share Posted October 31, 2023 Hello Everyone, Sorry I'm totally newbie with Frigate & NVR so probably my question/problem is really simple. I have Reolink Argus3 Cameras, and I'm tring to configure it with Frigate. I've installed Neolink, which is working without problems. Then I've installed Frigate and tested it with CPU as detector, and is working. Since I have a NVIDIA P4 on my Unraid server, I tried to configure it with NVidia, but I'm failing 🙂 The Frigate Docker is configured as following: And the tensorrt was configured : I modified the config.yml as following: mqtt: enabled: false cameras: Camera: ffmpeg: inputs: - path: rtsp://xxx.xxx.xxx.xxx:8554/xyz/main hwaccel_args: preset-nvidia-h264 roles: - detect - record objects: track: - person - dog - bicycle - car - motorcycle detect: width: 2560 height: 1440 record: enabled: True retain: days: 7 mode: all detectors: tensorrt: type: tensorrt device: 0 #This is the default, select the first GPU model: path: /trt-models/yolov7-tiny-416.trt input_tensor: nchw input_pixel_format: rgb width: 416 height: 416 I don't know what I'm missing here, since with this configuration, no imput is shown in Frigate. Quote Link to comment
yayitazale Posted November 2, 2023 Author Share Posted November 2, 2023 On 10/31/2023 at 5:18 PM, TDA said: Hello Everyone, Sorry I'm totally newbie with Frigate & NVR so probably my question/problem is really simple. I have Reolink Argus3 Cameras, and I'm tring to configure it with Frigate. I've installed Neolink, which is working without problems. Then I've installed Frigate and tested it with CPU as detector, and is working. Since I have a NVIDIA P4 on my Unraid server, I tried to configure it with NVidia, but I'm failing 🙂 The Frigate Docker is configured as following: And the tensorrt was configured : I modified the config.yml as following: mqtt: enabled: false cameras: Camera: ffmpeg: inputs: - path: rtsp://xxx.xxx.xxx.xxx:8554/xyz/main hwaccel_args: preset-nvidia-h264 roles: - detect - record objects: track: - person - dog - bicycle - car - motorcycle detect: width: 2560 height: 1440 record: enabled: True retain: days: 7 mode: all detectors: tensorrt: type: tensorrt device: 0 #This is the default, select the first GPU model: path: /trt-models/yolov7-tiny-416.trt input_tensor: nchw input_pixel_format: rgb width: 416 height: 416 I don't know what I'm missing here, since with this configuration, no imput is shown in Frigate. Did you add the "--runtime=nvidia" extra parameter? Quote Link to comment
bigthrilla Posted November 2, 2023 Share Posted November 2, 2023 On 10/31/2023 at 11:18 AM, TDA said: Hello Everyone, Sorry I'm totally newbie with Frigate & NVR so probably my question/problem is really simple. I have Reolink Argus3 Cameras, and I'm tring to configure it with Frigate. I've installed Neolink, which is working without problems. Then I've installed Frigate and tested it with CPU as detector, and is working. Since I have a NVIDIA P4 on my Unraid server, I tried to configure it with NVidia, but I'm failing 🙂 The Frigate Docker is configured as following: And the tensorrt was configured : I modified the config.yml as following: mqtt: enabled: false cameras: Camera: ffmpeg: inputs: - path: rtsp://xxx.xxx.xxx.xxx:8554/xyz/main hwaccel_args: preset-nvidia-h264 roles: - detect - record objects: track: - person - dog - bicycle - car - motorcycle detect: width: 2560 height: 1440 record: enabled: True retain: days: 7 mode: all detectors: tensorrt: type: tensorrt device: 0 #This is the default, select the first GPU model: path: /trt-models/yolov7-tiny-416.trt input_tensor: nchw input_pixel_format: rgb width: 416 height: 416 I don't know what I'm missing here, since with this configuration, no imput is shown in Frigate. TensorRT config is missing gpu ID from nvidia driver. Install nvidia driver on unraid and get ID from there Quote Link to comment
TDA Posted November 2, 2023 Share Posted November 2, 2023 (edited) 2 hours ago, yayitazale said: Did you add the "--runtime=nvidia" extra parameter? My bad, I was sure I had added it, but it was missing. Added now but same error. 1 hour ago, bigthrilla said: TensorRT config is missing gpu ID from nvidia driver. Install nvidia driver on unraid and get ID from there It blanked out in the printscreen, but it's inside. Is my config.yml correct? Particularly the camera config: preset-nvidia-h264 ? And the detectors are configured correctly? Do I have to set something particular inside neolink? Thanks to all in advance 🙂 EDIT: I outcommented: hwaccel_args: preset-nvidia-h264 And now it's "kind" of working. Not stable at the moment, and the playback of videos don't work. It's better to use GO2RTC in between? neolink --> go2rtc --> frigate with such cameras? Edited November 2, 2023 by TDA Added comment Quote Link to comment
yayitazale Posted November 2, 2023 Author Share Posted November 2, 2023 8 hours ago, TDA said: My bad, I was sure I had added it, but it was missing. Added now but same error. It blanked out in the printscreen, but it's inside. Is my config.yml correct? Particularly the camera config: preset-nvidia-h264 ? And the detectors are configured correctly? Do I have to set something particular inside neolink? Thanks to all in advance 🙂 EDIT: I outcommented: hwaccel_args: preset-nvidia-h264 And now it's "kind" of working. Not stable at the moment, and the playback of videos don't work. It's better to use GO2RTC in between? neolink --> go2rtc --> frigate with such cameras? Hi: Your config file doesn't look incorrect. However, you can add a Input Arg to the ffmpeg config. For reolink cameras is recommended yo use the http stream s input to the go2rtc restream and the use the restreamed feed to detection as you can read in the docs. Also, to be sure witch codecs are supported by your GPU, you should run this command inside the container terminal. Quote Link to comment
BurningSky Posted November 3, 2023 Share Posted November 3, 2023 I sent back the Coral USB I was having issues with and have swapped it with a mini PCIE one which I have connected to a PCIE to mPCIE adaptor but I am still having some issues. I have installed the drivers and it is being recognised by the Coral Driver app: Coral TPU1: Status: ALIVE Temperature: 39.30 °C Frequency: 500 MHz Driver Version: 1.2 Framework Version: 1.1.4 I can also see the device under sysdevs: [1ac1:089a] 25:00.0 System peripheral: Global Unichip Corp. Coral Edge TPU I had previously deleted the section for mapping the TPU but I have readded it with the following: Config Type: Device Name: Coral TPU/NCS2 Mapping Value: /dev/apex_0 Description: Use /dev/bus/usb for USB devices and /dev/apex_0 for PCIe devices (you must install the drivers first for PCIe devices). Remove this if you are not using it. In Frigate I added the following to detectors: detectors: coral: type: edgetpu device: pci But I get an error saying that no EdgeTPU was detected, have I misconfigured something? 2023-11-03 12:49:18.767219029 [INFO] Preparing go2rtc config... 2023-11-03 12:49:18.767580529 [INFO] Starting Frigate... 2023-11-03 12:49:18.768623682 [INFO] Starting NGINX... 2023-11-03 12:49:18.957519850 [WARN] Using go2rtc binary from '/config/go2rtc' instead of the embedded one 2023-11-03 12:49:18.960160839 [INFO] Starting go2rtc... 2023-11-03 12:49:19.063972373 12:49:19.063 INF go2rtc version 1.8.1 linux/amd64 2023-11-03 12:49:19.064282260 12:49:19.064 INF [api] listen addr=0.0.0.0:1984 2023-11-03 12:49:19.064866554 12:49:19.064 INF [rtsp] listen addr=0.0.0.0:8554 2023-11-03 12:49:19.064871792 12:49:19.064 INF [webrtc] listen addr=0.0.0.0:8555/tcp 2023-11-03 12:49:19.771751395 [2023-11-03 12:49:19] frigate.app INFO : Starting Frigate (0.12.1-367d724) 2023-11-03 12:49:19.811722914 [2023-11-03 12:49:19] peewee_migrate INFO : Starting migrations 2023-11-03 12:49:19.815743064 [2023-11-03 12:49:19] peewee_migrate INFO : There is nothing to migrate 2023-11-03 12:49:19.840242161 [2023-11-03 12:49:19] detector.coral INFO : Starting detection process: 577 2023-11-03 12:49:19.956163315 [2023-11-03 12:49:19] detector.cuda INFO : Starting detection process: 580 2023-11-03 12:49:19.956171347 [2023-11-03 12:49:19] frigate.app INFO : Output process started: 585 2023-11-03 12:49:19.956179867 [2023-11-03 12:49:19] frigate.detectors.plugins.edgetpu_tfl INFO : Attempting to load TPU as pci 2023-11-03 12:49:19.956187410 Process detector:coral: 2023-11-03 12:49:19.956194534 [2023-11-03 12:49:19] frigate.detectors.plugins.edgetpu_tfl INFO : TPU found 2023-11-03 12:49:19.956218629 [2023-11-03 12:49:19] frigate.detectors.plugins.edgetpu_tfl ERROR : No EdgeTPU was detected. If you do not have a Coral device yet, you must configure CPU detectors. Is there any way to check if the docker is actually seeing the device? Quote Link to comment
TDA Posted November 4, 2023 Share Posted November 4, 2023 On 11/2/2023 at 8:25 PM, yayitazale said: Hi: Your config file doesn't look incorrect. However, you can add a Input Arg to the ffmpeg config. For reolink cameras is recommended yo use the http stream s input to the go2rtc restream and the use the restreamed feed to detection as you can read in the docs. Also, to be sure witch codecs are supported by your GPU, you should run this command inside the container terminal. Hello, Thank you for the answer 🙂 I will search for a guide on how to configure go2rtc and try to configure it this way. Quote Link to comment
yayitazale Posted November 4, 2023 Author Share Posted November 4, 2023 8 hours ago, TDA said: Hello, Thank you for the answer 🙂 I will search for a guide on how to configure go2rtc and try to configure it this way. You have the info here: https://docs.frigate.video/guides/configuring_go2rtc Quote Link to comment
yayitazale Posted November 4, 2023 Author Share Posted November 4, 2023 On 11/3/2023 at 1:52 PM, BurningSky said: I sent back the Coral USB I was having issues with and have swapped it with a mini PCIE one which I have connected to a PCIE to mPCIE adaptor but I am still having some issues. I have installed the drivers and it is being recognised by the Coral Driver app: Coral TPU1: Status: ALIVE Temperature: 39.30 °C Frequency: 500 MHz Driver Version: 1.2 Framework Version: 1.1.4 I can also see the device under sysdevs: [1ac1:089a] 25:00.0 System peripheral: Global Unichip Corp. Coral Edge TPU I had previously deleted the section for mapping the TPU but I have readded it with the following: Config Type: Device Name: Coral TPU/NCS2 Mapping Value: /dev/apex_0 Description: Use /dev/bus/usb for USB devices and /dev/apex_0 for PCIe devices (you must install the drivers first for PCIe devices). Remove this if you are not using it. In Frigate I added the following to detectors: detectors: coral: type: edgetpu device: pci But I get an error saying that no EdgeTPU was detected, have I misconfigured something? 2023-11-03 12:49:18.767219029 [INFO] Preparing go2rtc config... 2023-11-03 12:49:18.767580529 [INFO] Starting Frigate... 2023-11-03 12:49:18.768623682 [INFO] Starting NGINX... 2023-11-03 12:49:18.957519850 [WARN] Using go2rtc binary from '/config/go2rtc' instead of the embedded one 2023-11-03 12:49:18.960160839 [INFO] Starting go2rtc... 2023-11-03 12:49:19.063972373 12:49:19.063 INF go2rtc version 1.8.1 linux/amd64 2023-11-03 12:49:19.064282260 12:49:19.064 INF [api] listen addr=0.0.0.0:1984 2023-11-03 12:49:19.064866554 12:49:19.064 INF [rtsp] listen addr=0.0.0.0:8554 2023-11-03 12:49:19.064871792 12:49:19.064 INF [webrtc] listen addr=0.0.0.0:8555/tcp 2023-11-03 12:49:19.771751395 [2023-11-03 12:49:19] frigate.app INFO : Starting Frigate (0.12.1-367d724) 2023-11-03 12:49:19.811722914 [2023-11-03 12:49:19] peewee_migrate INFO : Starting migrations 2023-11-03 12:49:19.815743064 [2023-11-03 12:49:19] peewee_migrate INFO : There is nothing to migrate 2023-11-03 12:49:19.840242161 [2023-11-03 12:49:19] detector.coral INFO : Starting detection process: 577 2023-11-03 12:49:19.956163315 [2023-11-03 12:49:19] detector.cuda INFO : Starting detection process: 580 2023-11-03 12:49:19.956171347 [2023-11-03 12:49:19] frigate.app INFO : Output process started: 585 2023-11-03 12:49:19.956179867 [2023-11-03 12:49:19] frigate.detectors.plugins.edgetpu_tfl INFO : Attempting to load TPU as pci 2023-11-03 12:49:19.956187410 Process detector:coral: 2023-11-03 12:49:19.956194534 [2023-11-03 12:49:19] frigate.detectors.plugins.edgetpu_tfl INFO : TPU found 2023-11-03 12:49:19.956218629 [2023-11-03 12:49:19] frigate.detectors.plugins.edgetpu_tfl ERROR : No EdgeTPU was detected. If you do not have a Coral device yet, you must configure CPU detectors. Is there any way to check if the docker is actually seeing the device? Are you running the container in privileged mode? Quote Link to comment
BurningSky Posted November 5, 2023 Share Posted November 5, 2023 14 hours ago, yayitazale said: Are you running the container in privileged mode? I didn't realise you couldn't have 2 types of detectors at the same time! Commented out the cuda one and it's working now! Quote Link to comment
noomis Posted November 5, 2023 Share Posted November 5, 2023 Hi, I have the following small problem with Frigate: I have set the retention period of recordings to 7 days and thought that one day at a time will be deleted. But unfortunately it is always deleted hour by hour and thus my array is rebooted once per hour. I have already tried to add expire_interval: 1440 in the config but unfortunately this does not work. I hope someone can help me 1 Quote Link to comment
reasonable-panel3162 Posted November 6, 2023 Share Posted November 6, 2023 (edited) Hi everyone, I'm totally new to Unraid and after spending 4 hours, I got almost things ready. I have a weird problem with Frigate and Nvidia at the moment. Frigate can see my GPU, but it won't use it for transcoding, both inside the container and in the host if I run nvidia-smi I can see it gives me the info of my GPU but there are no process are using it. In the frigate configuration, I have this: ffmpeg: hwaccel_args: preset-nvidia-h264 and to run my container with Nvidia GPU, I've added --runtime-nvidia and --gpus=all + two variables for gpu-id and capabilities. For a second I was happy it found my GPU but it's not using it --- Update 1 --- What's weird for me, when I remove the hwaccel_args the cpu usage goes up like 5-6 for a single camera, but when bring back the hwaccel_args the cpu usage goes down to fixed 2% Edited November 6, 2023 by reasonable-panel3162 Quote Link to comment
yayitazale Posted November 7, 2023 Author Share Posted November 7, 2023 On 11/5/2023 at 10:16 PM, noomis said: Hi, I have the following small problem with Frigate: I have set the retention period of recordings to 7 days and thought that one day at a time will be deleted. But unfortunately it is always deleted hour by hour and thus my array is rebooted once per hour. I have already tried to add expire_interval: 1440 in the config but unfortunately this does not work. I hope someone can help me You are totally right. I have monitored my server from yesterday to today and I have the exact same behavior. There is a open issue here so I have posted a question in the beta discussion witch is the version I'm using. https://github.com/blakeblackshear/frigate/discussions/8366#discussioncomment-7495998 https://github.com/blakeblackshear/frigate/issues/2365 lets see Quote Link to comment
yayitazale Posted November 7, 2023 Author Share Posted November 7, 2023 9 hours ago, reasonable-panel3162 said: Hi everyone, I'm totally new to Unraid and after spending 4 hours, I got almost things ready. I have a weird problem with Frigate and Nvidia at the moment. Frigate can see my GPU, but it won't use it for transcoding, both inside the container and in the host if I run nvidia-smi I can see it gives me the info of my GPU but there are no process are using it. In the frigate configuration, I have this: ffmpeg: hwaccel_args: preset-nvidia-h264 and to run my container with Nvidia GPU, I've added --runtime-nvidia and --gpus=all + two variables for gpu-id and capabilities. For a second I was happy it found my GPU but it's not using it --- Update 1 --- What's weird for me, when I remove the hwaccel_args the cpu usage goes up like 5-6 for a single camera, but when bring back the hwaccel_args the cpu usage goes down to fixed 2% First of all, if you added the gpu-id and the capabilities as env vars, you don't need to add --gpus=all in extra parameters, only the --runtime-nvidia one. You can run this into the container to test your GPUs codec compatibility: ffmpeg -decoders | grep cuvid 1 Quote Link to comment
reasonable-panel3162 Posted November 7, 2023 Share Posted November 7, 2023 11 minutes ago, yayitazale said: First of all, if you added the gpu-id and the capabilities as env vars, you don't need to add --gpus=all in extra parameters, only the --runtime-nvidia one. You can run this into the container to test your GPUs codec compatibility: ffmpeg -decoders | grep cuvid Hi, thanks for the reply. I'll remove gpus=all I already ran the ffmpeg -decoders and here is the result, it supports: libavutil 57. 28.100 / 57. 28.100 libavcodec 59. 37.100 / 59. 37.100 libavformat 59. 27.100 / 59. 27.100 libavdevice 59. 7.100 / 59. 7.100 libavfilter 8. 44.100 / 8. 44.100 libswscale 6. 7.100 / 6. 7.100 libswresample 4. 7.100 / 4. 7.100 libpostproc 56. 6.100 / 56. 6.100 V..... av1_cuvid Nvidia CUVID AV1 decoder (codec av1) V..... h264_cuvid Nvidia CUVID H264 decoder (codec h264) V..... hevc_cuvid Nvidia CUVID HEVC decoder (codec hevc) V..... mjpeg_cuvid Nvidia CUVID MJPEG decoder (codec mjpeg) V..... mpeg1_cuvid Nvidia CUVID MPEG1VIDEO decoder (codec mpeg1video) V..... mpeg2_cuvid Nvidia CUVID MPEG2VIDEO decoder (codec mpeg2video) V..... mpeg4_cuvid Nvidia CUVID MPEG4 decoder (codec mpeg4) V..... vc1_cuvid Nvidia CUVID VC1 decoder (codec vc1) V..... vp8_cuvid Nvidia CUVID VP8 decoder (codec vp8) V..... vp9_cuvid Nvidia CUVID VP9 decoder (codec vp9) Quote Link to comment
yayitazale Posted November 7, 2023 Author Share Posted November 7, 2023 5 hours ago, reasonable-panel3162 said: Hi, thanks for the reply. I'll remove gpus=all I already ran the ffmpeg -decoders and here is the result, it supports: libavutil 57. 28.100 / 57. 28.100 libavcodec 59. 37.100 / 59. 37.100 libavformat 59. 27.100 / 59. 27.100 libavdevice 59. 7.100 / 59. 7.100 libavfilter 8. 44.100 / 8. 44.100 libswscale 6. 7.100 / 6. 7.100 libswresample 4. 7.100 / 4. 7.100 libpostproc 56. 6.100 / 56. 6.100 V..... av1_cuvid Nvidia CUVID AV1 decoder (codec av1) V..... h264_cuvid Nvidia CUVID H264 decoder (codec h264) V..... hevc_cuvid Nvidia CUVID HEVC decoder (codec hevc) V..... mjpeg_cuvid Nvidia CUVID MJPEG decoder (codec mjpeg) V..... mpeg1_cuvid Nvidia CUVID MPEG1VIDEO decoder (codec mpeg1video) V..... mpeg2_cuvid Nvidia CUVID MPEG2VIDEO decoder (codec mpeg2video) V..... mpeg4_cuvid Nvidia CUVID MPEG4 decoder (codec mpeg4) V..... vc1_cuvid Nvidia CUVID VC1 decoder (codec vc1) V..... vp8_cuvid Nvidia CUVID VP8 decoder (codec vp8) V..... vp9_cuvid Nvidia CUVID VP9 decoder (codec vp9) So no idea what is going on... Do you see anything in the logs? I see this issue but I don't know if this is related or not: https://github.com/blakeblackshear/frigate/issues/8503 Quote Link to comment
reasonable-panel3162 Posted November 7, 2023 Share Posted November 7, 2023 4 hours ago, yayitazale said: So no idea what is going on... Do you see anything in the logs? I see this issue but I don't know if this is related or not: https://github.com/blakeblackshear/frigate/issues/8503 I can't also get my head over it why it's not working, there is nothing in nvidia-smi but from the CPU usage I can guess it's using GPU. If I disable hardware acceleration the CPU usage for a single camera goes above 5 percent, with that enabled it remains below 2%. it works better than fine with jellyfin for example, it transcode like hot sword on butter. Thanks for the help. Quote Link to comment
yayitazale Posted November 9, 2023 Author Share Posted November 9, 2023 On 11/7/2023 at 8:00 PM, reasonable-panel3162 said: I can't also get my head over it why it's not working, there is nothing in nvidia-smi but from the CPU usage I can guess it's using GPU. If I disable hardware acceleration the CPU usage for a single camera goes above 5 percent, with that enabled it remains below 2%. it works better than fine with jellyfin for example, it transcode like hot sword on butter. Thanks for the help. Are you launching the nvidia-smi inside the container or in the unraid terminal? Quote Link to comment
reasonable-panel3162 Posted November 9, 2023 Share Posted November 9, 2023 42 minutes ago, yayitazale said: Are you launching the nvidia-smi inside the container or in the unraid terminal? Both Quote Link to comment
yayitazale Posted November 9, 2023 Author Share Posted November 9, 2023 5 minutes ago, reasonable-panel3162 said: Both Can you post the entire config file and a screenshot of your docker deployment (advanced view)? You can hide the sensitive information. Quote Link to comment
reasonable-panel3162 Posted November 10, 2023 Share Posted November 10, 2023 17 hours ago, yayitazale said: Can you post the entire config file and a screenshot of your docker deployment (advanced view)? You can hide the sensitive information. I'm such an idiot, later I checked that frigate is using NVIDIA GPU and as a mistake I only run it inside the docker, I thought I also did outside of docker. Thanks and sorry. 1 Quote Link to comment
reasonable-panel3162 Posted November 12, 2023 Share Posted November 12, 2023 (edited) Frigate-nvr is consuming lots of docker-image, at the moment around 80GB. I ran df -H inside frigate container and this is the result: Filesystem Size Used Avail Use% Mounted on /dev/loop2 108G 83G 23G 79% / tmpfs 68M 0 68M 0% /dev shfs 2.0T 520G 1.5T 26% /frigate/media shm 68M 5.0M 63M 8% /dev/shm /dev/loop2 108G 83G 23G 79% /etc/hosts overlay 17G 1.3G 16G 8% /usr/bin/nvidia-smi rootfs 17G 1.3G 16G 8% /etc/vulkan/icd.d/nvidia_icd.json tmpfs 17G 13k 17G 1% /proc/driver/nvidia tmpfs 17G 4.1k 17G 1% /etc/nvidia/nvidia-application-profiles-rc.d overlay 17G 1.3G 16G 8% /lib/firmware/nvidia/545.29.02/gsp_ga10x.bin devtmpfs 8.4M 0 8.4M 0% /dev/nvidia0 Update 1: * To fix the issue at the moment, I recreated the image. * When I created it it was 8-9GB and now it's 13GB Edited November 12, 2023 by reasonable-panel3162 Quote Link to comment
tigga69 Posted November 12, 2023 Share Posted November 12, 2023 Just an FYI... the tensorrt_models.sh is out of date and doesnt work any more. Need the line to install packages changed to... # Install packages pip install --upgrade pip && pip install onnx==1.15.0 protobuf==4.25.0 Quote Link to comment
yayitazale Posted November 12, 2023 Author Share Posted November 12, 2023 11 hours ago, reasonable-panel3162 said: Frigate-nvr is consuming lots of docker-image, at the moment around 80GB. I ran df -H inside frigate container and this is the result: Filesystem Size Used Avail Use% Mounted on /dev/loop2 108G 83G 23G 79% / tmpfs 68M 0 68M 0% /dev shfs 2.0T 520G 1.5T 26% /frigate/media shm 68M 5.0M 63M 8% /dev/shm /dev/loop2 108G 83G 23G 79% /etc/hosts overlay 17G 1.3G 16G 8% /usr/bin/nvidia-smi rootfs 17G 1.3G 16G 8% /etc/vulkan/icd.d/nvidia_icd.json tmpfs 17G 13k 17G 1% /proc/driver/nvidia tmpfs 17G 4.1k 17G 1% /etc/nvidia/nvidia-application-profiles-rc.d overlay 17G 1.3G 16G 8% /lib/firmware/nvidia/545.29.02/gsp_ga10x.bin devtmpfs 8.4M 0 8.4M 0% /dev/nvidia0 Update 1: * To fix the issue at the moment, I recreated the image. * When I created it it was 8-9GB and now it's 13GB Can you post your configuration of the docker container? It looks you have deleted this from extra parameters: --mount type=tmpfs,target=/tmp/cache,tmpfs-size=1000000000 This the location for video recording segments caching as you can read here: https://docs.frigate.video/frigate/installation#storage Quote Link to comment
yayitazale Posted November 12, 2023 Author Share Posted November 12, 2023 (edited) 1 hour ago, tigga69 said: Just an FYI... the tensorrt_models.sh is out of date and doesnt work any more. Need the line to install packages changed to... # Install packages pip install --upgrade pip && pip install onnx==1.15.0 protobuf==4.25.0 In the next version witch is going to be released soon (currently the latest release is beta 5) the script (and also the container itself) is not going to be needed anymore to create the models: https://deploy-preview-6262--frigate-docs.netlify.app/configuration/object_detectors#generate-models Edited November 12, 2023 by yayitazale 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.