|
I installed Agent DVR and CodeProject.AI(v2.1.11-beta) Server on linux Using a camera.I registered the face name with Chinese in CodeProject.AI Explorer. But the Chinese label is displayed as panes when face recognition.How to set CodeProject.AI.
Thank you.
|
|
|
|
|
John Martin 2023 wrote: the Chinese label is displayed as panes when face recognition.
Where are you seeing this? AgentDVR or CodeProject.AI? Can you let open the explorer, go to the Face panel and hit "list registered faces" and let me know if the problem is there?
cheers
Chris Maunder
|
|
|
|
|
Hello,
I'm running the CPAI with the ObjectDetection module (Coral) in docker.
It's working, but I would like to run google coral with maximum performance, how do I force the installation of the lib
libedgetpu1-max
modified 22-Aug-23 9:32am.
|
|
|
|
|
Stop the Coral module from the CPAI web interface.
Get into the Docker container’s shell. (docker exec codeprojectai /bin/bash)
Do “sudo apt-get remove libedgetpu1-std”.
Do “DEBIAN_FRONTEND=dialog sudo apt-get install libedgetpu1-max”.
Start the Coral module from the CPAI web interface.
Done. As long as you don’t pull a new image or do something else that recreates the container (simply stopping and restarting is fine), this should stick.
|
|
|
|
|
I can start codeproject and it will run for days, but as soon as I start blue iris or select "Use AI on port ##" in blue iris, codeproject will shut down, and blue iris will become very sluggish.
It was working good for a couple months, but when I updated BlueIris and Codeproject, it started acting crazy.
I have installed Codeproject on a different machine and still get the same results.
I have totally uninstalled Codeproject and delete the folders and reinstalled, but still no joy!
|
|
|
|
|
Thanks very much for the report. Do you have auto start / stop with start with Blue Iris checked off? Try unchecking that box if it is checked.
Thanks,
Sean Ewington
CodeProject
|
|
|
|
|
That seems to have fixed it, so simple,
Thanks!
|
|
|
|
|
Hoping this is the right place to post this.
I am running codeproject.ai in a docker container on an ubuntu 22.04 machine. The machine has 2 nvidia tesla k80's installed. Both cards are functioning properly. I am currently using codeproject.ai with Blue Iris and everything is running fine but it only works with CPU and not the 4 GPU's.
I cannot figure out where the issue is and how to get codeproject to use the gpu's rather than the cpu.
What I have tried:
I have ensured that nvidia-smi works from both the machine and container.
I have ensured that nvidia is the default runtime for docker.
I have ensured that all gpu's are accessible to the container.
I have tried enabling the gpu for each module.
I have stopped and restarted the modules.
I have removed the container and persistent volumes multiple times.
Here is my docker run command
sudo docker run --name CodeProject.AI -d -p 32168:32168 \
--gpus all \
--mount type=bind,source=/home/dax/docker/codeprojectai/ai,target=/etc/codeproject/ai \
--mount type=bind,source=/home/dax/docker/codeprojectai/modules,target=/app/modules \
--mount type=bind,source=/home/dax/docker/codeprojectai/custom-modules,target=/app/modules/ObjectDetectionYolo/custom-models \
codeproject/ai-server:gpu
|
|
|
|
|
Thanks very much for the report. Does CPAI work in the CPAI Explorer?
Thanks,
Sean Ewington
CodeProject
|
|
|
|
|
Pardon my ignorance, but what do you mean by the CPAI Explorer?
Edit: I found it. I'll update shortly.
|
|
|
|
|
Can you please post the contents of the System Info tab? I'm wondering if the issue is that you may need to explicitly set the device name in the modulesettings.json file. (Just clutching at straws because you seem to have done everything I would have suggested)
cheers
Chris Maunder
|
|
|
|
|
Server version: 2.1.11-Beta
Operating System: Linux (Linux 6.2.0-26-generic #26~22.04.1-Ubuntu SMP PREEMPT_DYNAMIC Thu Jul 13 16:27:29 UTC 2)
CPUs: 12th Gen Intel(R) Core(TM) i7-12700K (Intel)
1 CPU x 12 cores. 20 logical processors (x64)
GPU: (NVIDIA) CUDA: (capable: ) Compute:
System RAM: 63 GiB
Target: Linux
BuildConfig: Release
Execution Env: Docker
Runtime Env: Production
.NET framework: .NET 7.0.10
System GPU info:
GPU 3D Usage 0%
GPU RAM Usage 0
Video adapter info:
Global Environment variables:
CPAI_APPROOTPATH = /app
CPAI_PORT = 32168
|
|
|
|
|
That's interesting: No NVIDIA info is showing.
Can you do me a favour and open a terminal into the docker container and type:
nvidia-smi and also nvcc --version and let me know what it says?
cheers
Chris Maunder
|
|
|
|
|
Hello,
I know CodeProject.AI is supported in blueiris and agentnvr and other systems but they just arent it for me.
I wanted to know if its possible to integrate with DigitalWatchdog Spectrum or preferably Hanwha Wave.
If it is possible could this be added with a bounty?
|
|
|
|
|
|
|
Hi,
I've purchased a Coral Edge Dual TPU module along with an M.2 Adaptor. I know someone who successfully is running one of the edge dual models on a PCIE adaptor, so no reason why this shouldn't work. But it won't launch.
Here it becomes a bit blurry - there's no guide for CPAI on how to install the various Coral Adaptors - an oversight Code Project?? The Coral Guide suggests a series of steps, but I have reason to believe some steps probably aren't necessary with CPAI (stand to be corrected).
Steps so far:
1. Physically installed the Coral Edge Dual & Adaptor
2. Installed the Visual Studio Re distributable from the Microsoft Website (This should have installed the 2022 version)
3. Installed the EdgeTpuRuntime Package that installs the Coral TPU driver and the Edge TPU Runtime software.
Upon starting BI / CPAI, I found that Yolo 5 6.2 was being reported as stopped despite me having removed it. I've also found that the Coral TPU adpator isn't being started and is exiting.
Can anyone at CPAI please help with this?
These are the logs:
Quote: An update to version 2.1.11-Beta is available Download
Improvements to YOLO model training.
09:00:43:Operating System: Windows (Microsoft Windows 10.0.19045)
09:00:43:CPUs: 11th Gen Intel(R) Core(TM) i5-11400 @ 2.60GHz (Intel)
09:00:43: 1 CPU x 6 cores. 12 logical processors (x64)
09:00:43:GPU: Intel(R) UHD Graphics 730 (1,024 MiB) (Intel Corporation)
09:00:43: Driver: 30.0.101.1273
09:00:43:System RAM: 16 GiB
09:00:43:Target: Windows
09:00:43:BuildConfig: Release
09:00:43:Execution Env: Native
09:00:43:Runtime Env: Production
09:00:43:.NET framework: .NET 7.0.5
09:00:43:App DataDir: C:\ProgramData\CodeProject\AI
09:00:43:Video adapter info:
09:00:43: Intel(R) UHD Graphics 730:
09:00:43: Driver Version 30.0.101.1273
09:00:43: Video Processor Intel(R) UHD Graphics Family
09:00:43:ROOT_PATH = C:\Program Files\CodeProject\AI
09:00:43:RUNTIMES_PATH = C:\Program Files\CodeProject\AI\runtimes
09:00:43:PREINSTALLED_MODULES_PATH = C:\Program Files\CodeProject\AI\preinstalled-modules
09:00:43:MODULES_PATH = C:\Program Files\CodeProject\AI\modules
09:00:43:PYTHON_PATH = \bin\windows\%PYTHON_RUNTIME%\venv\scripts\Python
09:00:43:Data Dir = C:\ProgramData\CodeProject\AI
09:00:43:Server version: 2.1.10-Beta
09:00:43:ModuleRunner Start
09:00:43:Starting Background AI Modules
09:00:45:Client request 'list-custom' in queue 'objectdetection_queue' (...321197)
09:00:46:GetCommandByRuntime: Runtime=python37, Location=Shared
09:00:46:Command: C:\Program Files\CodeProject\AI\runtimes\bin\windows\python37\venv\scripts\Python
09:00:46:
09:00:46:Starting C:\Program Files...ws\python37\venv\scripts\Python "C:\Program Files...tectionYolo\detect_adapter.py"
09:00:46:Attempting to start ObjectDetectionYolo with C:\Program Files\CodeProject\AI\runtimes\bin\windows\python37\venv\scripts\Python "C:\Program Files\CodeProject\AI\modules\ObjectDetectionYolo\detect_adapter.py"
09:00:46:
09:00:46:Module 'Object Detection (YOLOv5 6.2)' (ID: ObjectDetectionYolo)
09:00:46:Module Path: C:\Program Files\CodeProject\AI\modules\ObjectDetectionYolo
09:00:46:AutoStart: True
09:00:46:Queue: objectdetection_queue
09:00:46:Platforms: all
09:00:46:GPU: Support enabled
09:00:46:Parallelism: 0
09:00:46:Accelerator:
09:00:46:Half Precis.: enable
09:00:46:Runtime: python37
09:00:46:Runtime Loc: Shared
09:00:46:FilePath: detect_adapter.py
09:00:46:Pre installed: False
09:00:46:Start pause: 1 sec
09:00:46:LogVerbosity:
09:00:46:Valid: True
09:00:46:Environment Variables
09:00:46:APPDIR = %CURRENT_MODULE_PATH%
09:00:46:CPAI_MODULE_SUPPORT_GPU = False
09:00:46:CUSTOM_MODELS_DIR = %CURRENT_MODULE_PATH%/custom-models
09:00:46:MODELS_DIR = %CURRENT_MODULE_PATH%/assets
09:00:46:MODEL_SIZE = Medium
09:00:46:USE_CUDA = False
09:00:46:YOLOv5_AUTOINSTALL = false
09:00:46:YOLOv5_VERBOSE = false
09:00:46:
09:00:46:Started Object Detection (YOLOv5 6.2) module
09:00:48:A new version 2.1.11-Beta is available
09:00:48:Current Version is 2.1.10-Beta
09:00:49:detect_adapter.py: CPAI_MODULE_REQUIRED_MB not found. Setting to default 0
09:00:49:detect_adapter.py: APPDIR: C:\Program Files\CodeProject\AI\modules\ObjectDetectionYolo
09:00:49:detect_adapter.py: MODEL_SIZE: medium
09:00:49:detect_adapter.py: MODELS_DIR: C:\Program Files\CodeProject\AI\modules\ObjectDetectionYolo\assets
09:00:49:Request 'list-custom' dequeued from 'objectdetection_queue' (...321197)
09:00:49:Running init for Object Detection (YOLOv5 6.2)
09:00:49:Object Detection (YOLOv5 6.2): Object Detection (YOLOv5 6.2) started.
09:00:49:Object Detection (YOLOv5 6.2): Retrieved objectdetection_queue command
09:00:49:Object Detection (YOLOv5 6.2): Queue request for Object Detection (YOLOv5 6.2) command 'list-custom' (...321197) took 1ms
09:00:49:Response received (...321197)
09:00:50:Sending shutdown request to python/ObjectDetectionYolo
09:00:50:Client request 'Quit' in queue 'objectdetection_queue' (...0c7f7f)
09:00:50:Request 'Quit' dequeued from 'objectdetection_queue' (...0c7f7f)
09:00:50:Object Detection (YOLOv5 6.2): Retrieved objectdetection_queue command
09:00:59:Module ObjectDetectionYolo has shutdown
09:00:59:detect_adapter.py: has exited
09:01:20:Preparing to install module 'ObjectDetectionCoral'
09:01:20:Downloading module 'ObjectDetectionCoral'
09:01:21:Installing module 'ObjectDetectionCoral'
09:01:21:ObjectDetectionCoral: Installing CodeProject.AI Analysis Module
09:01:21:ObjectDetectionCoral: ========================================================================
09:01:21:ObjectDetectionCoral: CodeProject.AI Installer
09:01:21:ObjectDetectionCoral: ========================================================================
09:01:21:ObjectDetectionCoral: Checking GPU support
09:01:21:ObjectDetectionCoral: CUDA Present...False
09:01:21:ObjectDetectionCoral: Allowing GPU Support: Yes
09:01:21:ObjectDetectionCoral: Allowing CUDA Support: No
09:01:21:ObjectDetectionCoral: General CodeProject.AI setup
09:01:21:ObjectDetectionCoral: Creating Directories...Done
09:01:21:ObjectDetectionCoral: Installing module ObjectDetectionCoral
09:01:21:ObjectDetectionCoral: Installing python37 in C:\Program Files\CodeProject\AI\runtimes\bin\windows\python37
09:01:21:ObjectDetectionCoral: Checking for python37 download...Present
09:01:23:ObjectDetectionYolo went quietly
09:01:23:GetCommandByRuntime: Runtime=python37, Location=Shared
09:01:23:Command: C:\Program Files\CodeProject\AI\runtimes\bin\windows\python37\venv\scripts\Python
09:01:23:Starting C:\Program Files...ws\python37\venv\scripts\Python "C:\Program Files...tectionYolo\detect_adapter.py"
09:01:23:
09:01:23:Attempting to start ObjectDetectionYolo with C:\Program Files\CodeProject\AI\runtimes\bin\windows\python37\venv\scripts\Python "C:\Program Files\CodeProject\AI\modules\ObjectDetectionYolo\detect_adapter.py"
09:01:23:
09:01:23:Module 'Object Detection (YOLOv5 6.2)' (ID: ObjectDetectionYolo)
09:01:23:Module Path: C:\Program Files\CodeProject\AI\modules\ObjectDetectionYolo
09:01:23:AutoStart: True
09:01:23:Queue: objectdetection_queue
09:01:23:Platforms: all
09:01:23:GPU: Support enabled
09:01:23:Parallelism: 0
09:01:23:Accelerator:
09:01:23:Half Precis.: enable
09:01:23:Runtime: python37
09:01:23:Runtime Loc: Shared
09:01:23:FilePath: detect_adapter.py
09:01:23:Pre installed: False
09:01:23:Start pause: 1 sec
09:01:23:LogVerbosity:
09:01:23:Valid: True
09:01:23:Environment Variables
09:01:23:APPDIR = %CURRENT_MODULE_PATH%
09:01:23:CPAI_MODULE_SUPPORT_GPU = False
09:01:23:CUSTOM_MODELS_DIR = %CURRENT_MODULE_PATH%/custom-models
09:01:23:MODELS_DIR = %CURRENT_MODULE_PATH%/assets
09:01:23:MODEL_SIZE = Medium
09:01:23:USE_CUDA = False
09:01:23:YOLOv5_AUTOINSTALL = false
09:01:23:YOLOv5_VERBOSE = false
09:01:23:
09:01:23:Started Object Detection (YOLOv5 6.2) module
09:01:26:detect_adapter.py: CPAI_MODULE_REQUIRED_MB not found. Setting to default 0
09:01:26:detect_adapter.py: APPDIR: C:\Program Files\CodeProject\AI\modules\ObjectDetectionYolo
09:01:26:detect_adapter.py: MODEL_SIZE: medium
09:01:26:detect_adapter.py: MODELS_DIR: C:\Program Files\CodeProject\AI\modules\ObjectDetectionYolo\assets
09:01:26:Running init for Object Detection (YOLOv5 6.2)
09:01:26:Object Detection (YOLOv5 6.2): Object Detection (YOLOv5 6.2) started.
09:01:27:ObjectDetectionCoral: Creating Virtual Environment...Done
09:01:27:ObjectDetectionCoral: Enabling our Virtual Environment...Done
09:01:27:ObjectDetectionCoral: Confirming we have Python 3.7...present
09:01:28:ObjectDetectionCoral: Ensuring Python package manager (pip) is installed...Done
09:01:33:ObjectDetectionCoral: Ensuring Python package manager (pip) is up to date...Done
09:01:33:ObjectDetectionCoral: Choosing Python packages from requirements.txt
09:01:38:ObjectDetectionCoral: Installing Packages into Virtual Environment...Success
09:01:39:ObjectDetectionCoral: Ensuring Python package manager (pip) is installed...Done
09:01:39:ObjectDetectionCoral: Ensuring Python package manager (pip) is up to date...Done
09:01:39:ObjectDetectionCoral: Choosing Python packages from requirements.txt
09:01:46:ObjectDetectionCoral: Installing Packages into Virtual Environment...Success
09:01:50:ObjectDetectionCoral: Downloading edge TPU runtime...Expanding...Done.
09:01:50:ObjectDetectionCoral: *** You may need to run C:\Program Files\CodeProject\AI\modules\ObjectDetectionCoral\edgetpu_runtime\install.bat to complete this process. Attempting to run this script now.
09:01:50:ObjectDetectionCoral: Installing UsbDk
09:01:50:ObjectDetectionCoral: Installing Windows drivers
09:01:50:ObjectDetectionCoral: Microsoft PnP Utility
09:01:50:ObjectDetectionCoral: Adding driver package: coral.inf
09:01:50:ObjectDetectionCoral: Driver package added successfully. (Already exists in the system)
09:01:50:ObjectDetectionCoral: Published Name: oem10.inf
09:01:50:ObjectDetectionCoral: Driver package installed on device: PCI\VEN_1AC1&DEV_089A&SUBSYS_089A1AC1&REV_00\6&fb5d63b&0&00380030
09:01:50:ObjectDetectionCoral: Driver package installed on device: PCI\VEN_1AC1&DEV_089A&SUBSYS_089A1AC1&REV_00\6&1b005fa9&0&00180030
09:01:50:ObjectDetectionCoral: Adding driver package: Coral_USB_Accelerator.inf
09:01:50:ObjectDetectionCoral: Driver package added successfully. (Already exists in the system)
09:01:50:ObjectDetectionCoral: Published Name: oem14.inf
09:01:50:ObjectDetectionCoral: Adding driver package: Coral_USB_Accelerator_(DFU).inf
09:01:50:ObjectDetectionCoral: Driver package added successfully. (Already exists in the system)
09:01:50:ObjectDetectionCoral: Published Name: oem16.inf
09:01:50:ObjectDetectionCoral: Total driver packages: 3
09:01:50:ObjectDetectionCoral: Added driver packages: 3
09:01:50:ObjectDetectionCoral: Installing performance counters
09:01:50:ObjectDetectionCoral: Info: Provider {aaa5bf9e-c44b-4177-af65-d3a06ba45fe7} defined in C:\Program Files\CodeProject\AI\modules\ObjectDetectionCoral\edgetpu_runtime\third_party\coral_accelerator_windows\coral.man is already installed in system repository.
09:01:50:ObjectDetectionCoral: Info: Successfully installed performance counters in C:\Program Files\CodeProject\AI\modules\ObjectDetectionCoral\edgetpu_runtime\third_party\coral_accelerator_windows\coral.manCopying edgetpu and libusb to System32
09:01:50:ObjectDetectionCoral: 1 file(s) copied.
09:01:51:ObjectDetectionCoral: 1 file(s) copied.
09:01:51:ObjectDetectionCoral: Install complete
09:02:07:ObjectDetectionCoral: Downloading MobileNet models...Expanding...Done.
09:02:07:ObjectDetectionCoral: Module setup complete
09:02:07:Module ObjectDetectionCoral installed successfully.
09:02:07:Module ObjectDetectionCoral not configured to AutoStart.
09:02:07:Installer exited with code 0
09:03:24:Client request 'list-custom' in queue 'objectdetection_queue' (...7f387a)
09:03:24:Request 'list-custom' dequeued from 'objectdetection_queue' (...7f387a)
09:03:24:Response received (...7f387a)
09:03:24:Object Detection (YOLOv5 6.2): Retrieved objectdetection_queue command
09:03:24:Object Detection (YOLOv5 6.2): Queue request for Object Detection (YOLOv5 6.2) command 'list-custom' (...7f387a) took 1ms
09:03:25:Sending shutdown request to python/ObjectDetectionYolo
09:03:25:Client request 'Quit' in queue 'objectdetection_queue' (...3e58ba)
09:03:25:Request 'Quit' dequeued from 'objectdetection_queue' (...3e58ba)
09:03:25:Object Detection (YOLOv5 6.2): Retrieved objectdetection_queue command
09:03:34:Module ObjectDetectionYolo has shutdown
09:03:34:detect_adapter.py: has exited
09:03:58:ObjectDetectionYolo went quietly
09:03:58:GetCommandByRuntime: Runtime=python37, Location=Shared
09:03:58:Command: C:\Program Files\CodeProject\AI\runtimes\bin\windows\python37\venv\scripts\Python
09:03:58:
09:03:58:Attempting to start ObjectDetectionYolo with C:\Program Files\CodeProject\AI\runtimes\bin\windows\python37\venv\scripts\Python "C:\Program Files\CodeProject\AI\modules\ObjectDetectionYolo\detect_adapter.py"
09:03:58:
09:03:58:Module 'Object Detection (YOLOv5 6.2)' (ID: ObjectDetectionYolo)
09:03:58:Module Path: C:\Program Files\CodeProject\AI\modules\ObjectDetectionYolo
09:03:58:AutoStart: True
09:03:58:Queue: objectdetection_queue
09:03:58:Platforms: all
09:03:58:GPU: Support enabled
09:03:58:Parallelism: 0
09:03:58:Starting C:\Program Files...ws\python37\venv\scripts\Python "C:\Program Files...tectionYolo\detect_adapter.py"
09:03:58:Accelerator:
09:03:58:Half Precis.: enable
09:03:58:Runtime: python37
09:03:58:Runtime Loc: Shared
09:03:58:FilePath: detect_adapter.py
09:03:58:Pre installed: False
09:03:58:Start pause: 1 sec
09:03:58:LogVerbosity:
09:03:58:Valid: True
09:03:58:Environment Variables
09:03:58:APPDIR = %CURRENT_MODULE_PATH%
09:03:58:CPAI_MODULE_SUPPORT_GPU = False
09:03:58:CUSTOM_MODELS_DIR = %CURRENT_MODULE_PATH%/custom-models
09:03:58:MODELS_DIR = %CURRENT_MODULE_PATH%/assets
09:03:58:MODEL_SIZE = Medium
09:03:58:USE_CUDA = False
09:03:58:YOLOv5_AUTOINSTALL = false
09:03:58:YOLOv5_VERBOSE = false
09:03:58:
09:03:58:Started Object Detection (YOLOv5 6.2) module
09:04:01:detect_adapter.py: CPAI_MODULE_REQUIRED_MB not found. Setting to default 0
09:04:01:detect_adapter.py: APPDIR: C:\Program Files\CodeProject\AI\modules\ObjectDetectionYolo
09:04:01:detect_adapter.py: MODEL_SIZE: medium
09:04:01:detect_adapter.py: MODELS_DIR: C:\Program Files\CodeProject\AI\modules\ObjectDetectionYolo\assets
09:04:01:Running init for Object Detection (YOLOv5 6.2)
09:04:01:Object Detection (YOLOv5 6.2): Object Detection (YOLOv5 6.2) started.
09:04:41:Sending shutdown request to python/ObjectDetectionYolo
09:04:41:Client request 'Quit' in queue 'objectdetection_queue' (...eda207)
09:04:41:Request 'Quit' dequeued from 'objectdetection_queue' (...eda207)
09:04:41:Object Detection (YOLOv5 6.2): Retrieved objectdetection_queue command
09:04:51:Module ObjectDetectionYolo has shutdown
09:04:51:detect_adapter.py: has exited
09:05:14:ObjectDetectionYolo went quietly
09:07:27:Client request 'list-custom' in queue 'objectdetection_queue' (...fa36bf)
09:10:13:Client request 'list-custom' in queue 'objectdetection_queue' (...e10fa8)
Coral suggests installing a python Library (complex set of command line options and seems aimed at teaching / developing a module) and TensorFlow something or other, but I rather think these are probably elements contained within the CPAI module, or at least they should be.
Anyone at CPAI help with this?
|
|
|
|
|
Our installer script does (or attempts to do) all those painful complex steps all the guides suggest. We're trying hard to take those 10 different ways to do something and automate it all so you don't have to go through the pain of what we've gone through to get things working.
It looks like it all installed correctly. On the dashboard can you click the "..." to the right of the ObjectDetection YOLO package and select Stop. Click the ... next to the Coral module and select Start. Do the modules appear in the correct state?
If so, open the Explorer (link is top of dashboard in blue) and test the object detection through the explorer UI
cheers
Chris Maunder
|
|
|
|
|
Hi Chris,
Thanks for the reply.
I tried starting it from the "...." and this time it started. However, some errors were seen.
The logs are below. It also seemed to decide no settings had been set and set it's own even though there were already settings in the dashboard. I've had no triggers so hard to say if it's working atm. It has de-conflicted the previous triggers, but there's no sign of them being processed or processing times.
So anyone's guess if it's actually working or part working.
The error messages can be seen in the logs below:
20:01:28:Operating System: Windows (Microsoft Windows 10.0.19045)
20:01:28:CPUs: 11th Gen Intel(R) Core(TM) i5-11400 @ 2.60GHz (Intel)
20:01:28: 1 CPU x 6 cores. 12 logical processors (x64)
20:01:28:GPU: Intel(R) UHD Graphics 730 (1,024 MiB) (Intel Corporation)
20:01:28: Driver: 30.0.101.1273
20:01:28:System RAM: 16 GiB
20:01:28:Target: Windows
20:01:28:BuildConfig: Release
20:01:28:Execution Env: Native
20:01:28:Runtime Env: Production
20:01:28:.NET framework: .NET 7.0.5
20:01:28:App DataDir: C:\ProgramData\CodeProject\AI
20:01:28:Video adapter info:
20:01:28: Intel(R) UHD Graphics 730:
20:01:28: Driver Version 30.0.101.1273
20:01:28: Video Processor Intel(R) UHD Graphics Family
20:01:28:ROOT_PATH = C:\Program Files\CodeProject\AI
20:01:28:RUNTIMES_PATH = C:\Program Files\CodeProject\AI\runtimes
20:01:28:PREINSTALLED_MODULES_PATH = C:\Program Files\CodeProject\AI\preinstalled-modules
20:01:28:MODULES_PATH = C:\Program Files\CodeProject\AI\modules
20:01:28:PYTHON_PATH = \bin\windows\%PYTHON_RUNTIME%\venv\scripts\Python
20:01:28:Data Dir = C:\ProgramData\CodeProject\AI
20:01:28:Server version: 2.1.10-Beta
20:01:28:ModuleRunner Start
20:01:28:Starting Background AI Modules
20:01:33:A new version 2.1.11-Beta is available
20:01:33:Current Version is 2.1.10-Beta
20:01:39:Client request 'detect' in queue 'objectdetection_queue' (...31f90a)
20:01:39:Client request 'detect' in queue 'objectdetection_queue' (...761508)
20:01:40:Client request 'detect' in queue 'objectdetection_queue' (...7a1c02)
20:01:57:Client request 'detect' in queue 'objectdetection_queue' (...bb3459)
20:01:57:Client request 'detect' in queue 'objectdetection_queue' (...681ef6)
20:01:58:Client request 'detect' in queue 'objectdetection_queue' (...245226)
23:03:55:GetCommandByRuntime: Runtime=python37, Location=Local
23:03:55:Command: C:\Program Files\CodeProject\AI\modules\ObjectDetectionCoral\bin\windows\python37\venv\scripts\Python
23:03:55:
23:03:55:
23:03:55:Module 'ObjectDetection (Coral)' (ID: ObjectDetectionCoral)
23:03:55:Module Path: C:\Program Files\CodeProject\AI\modules\ObjectDetectionCoral
23:03:55:Starting C:\Program Files...ws\python37\venv\scripts\Python "C:\Program Files...ectdetection_coral_adapter.py"
23:03:55:Attempting to start ObjectDetectionCoral with C:\Program Files\CodeProject\AI\modules\ObjectDetectionCoral\bin\windows\python37\venv\scripts\Python "C:\Program Files\CodeProject\AI\modules\ObjectDetectionCoral\objectdetection_coral_adapter.py"
23:03:55:AutoStart: True
23:03:55:Queue: objectdetection_queue
23:03:55:Platforms: windows,linux,linux-arm64,macos,macos-arm64
23:03:55:GPU: Support enabled
23:03:55:Parallelism: 1
23:03:55:Accelerator:
23:03:55:Half Precis.: enable
23:03:55:Runtime: python37
23:03:55:Runtime Loc: Local
23:03:55:FilePath: objectdetection_coral_adapter.py
23:03:55:Pre installed: False
23:03:55:Start pause: 1 sec
23:03:55:LogVerbosity:
23:03:55:Valid: True
23:03:55:Environment Variables
23:03:55:MODELS_DIR = %CURRENT_MODULE_PATH%/assets
23:03:55:MODEL_SIZE = Medium
23:03:55:
23:03:55:Started ObjectDetection (Coral) module
23:04:11:objectdetection_coral_adapter.py: E driver/mmio_driver.cc:254] HIB Error. hib_error_status = ffffffffffffffff, hib_first_error_status = ffffffffffffffff
23:04:24:objectdetection_coral_adapter.py: E driver/mmio_driver.cc:254] HIB Error. hib_error_status = ffffffffffffffff, hib_first_error_status = ffffffffffffffff
23:04:24:objectdetection_coral_adapter.py: CPAI_MODULE_REQUIRED_MB not found. Setting to default 0
23:04:24:objectdetection_coral_adapter.py: NUM_THREADS not found. Setting to default 1
23:04:24:objectdetection_coral_adapter.py: MIN_CONFIDENCE not found. Setting to default 0.5
23:04:24:objectdetection_coral_adapter.py: MODULE_PATH: C:\Program Files\CodeProject\AI\modules\ObjectDetectionCoral
23:04:24:objectdetection_coral_adapter.py: MODELS_DIR: C:\Program Files\CodeProject\AI\modules\ObjectDetectionCoral\assets
23:04:24:objectdetection_coral_adapter.py: MODEL_SIZE: medium
23:04:24:objectdetection_coral_adapter.py: Timeout connecting to the server
23:04:24:objectdetection_coral_adapter.py: ObjectDetection (Coral) started.ObjectDetection (Coral): ObjectDetection (Coral) started.
23:04:24:objectdetection_coral_adapter.py: CPU_MODEL_NAME: efficientdet_lite3_512_ptq.tflite
23:04:24:objectdetection_coral_adapter.py: TPU_MODEL_NAME: efficientdet_lite3_512_ptq_edgetpu.tflite
23:04:24:objectdetection_coral_adapter.py: Input details: {'name': 'serving_default_images ', 'index': 0, 'shape': array([ 1, 512, 512, 3]), 'shape_signature': array([ 1, 512, 512, 3]), 'dtype': , 'quantization': (0.0078125, 127), 'quantization_parameters': {'scales': array([0.0078125], dtype=float32), 'zero_points': array([127]), 'quantized_dimension': 0}, 'sparsity_parameters': {}}
23:04:24:objectdetection_coral_adapter.py: Output details: {'name': 'StatefulPartitionedCall:31', 'index': 931, 'shape': array([ 1, 25, 4]), 'shape_signature': array([ 1, 25, 4]), 'dtype': , 'quantization': (0.0, 0), 'quantization_parameters': {'scales': array([], dtype=float32), 'zero_points': array([], dtype=int32), 'quantized_dimension': 0}, 'sparsity_parameters': {}}
23:04:24:ObjectDetection (Coral): ObjectDetection (Coral) started.
23:04:24:Request 'detect' dequeued from 'objectdetection_queue' (...31f90a)
23:04:24:Request 'detect' dequeued from 'objectdetection_queue' (...761508)
23:04:24:Request 'detect' dequeued from 'objectdetection_queue' (...7a1c02)
23:04:24:Request 'detect' dequeued from 'objectdetection_queue' (...681ef6)
23:04:24:Request 'detect' dequeued from 'objectdetection_queue' (...bb3459)
23:04:24:Request 'detect' dequeued from 'objectdetection_queue' (...245226)
One further point. I was unsure if I needed to load custom models for the EdgeTPU so stopped the service as it wouldn't allow custom models without. I was unable to select anything. BI then hung for quite some time before coming back and starting the AI service again. Seems to still be some code conflict there between AI and BI.
|
|
|
|
|
Ghost Man Nov2022 wrote: It also seemed to decide no settings had been set and set it's own even though there were already settings in the dashboard.
Do you mean CodeProject settings or Blue Iris settings?
Ghost Man Nov2022 wrote: I've had no triggers so hard to say if it's working atm. It has de-conflicted the previous triggers, but there's no sign of them being processed or processing times.
I assume this is in Blue Iris
Ghost Man Nov2022 wrote: I was unsure if I needed to load custom models for the EdgeTPU
We don't have custom models for the Coral yet, but will certainly be adding that soon.
In the logs I'm not seeing that the TPU was actually detected. Can you check your hardware?
cheers
Chris Maunder
|
|
|
|
|
Quote: Do you mean CodeProject settings or Blue Iris settings?
The C{AI logs suggested the values for minimum confidence etc were blank but in the AI Tab for BI they were filled out:
Quote: I assume this is in Blue Iris
Yes.
Quote: We don't have custom models for the Coral yet, but will certainly be adding that soon.
Ok. I was just wondering if I had to select a Coral model for it to work.
|
|
|
|
|
Quote: In the logs I'm not seeing that the TPU was actually detected. Can you check your hardware?
2 instances reported (Dual TPU) and windows reports it's working properly, so it's at least seeing it and has drivers.
I know someone with one working and they manually installed both Python using a script and Tensor flow lite, so I'm guessing that the issue is the CPAI module isn't correctly installing one of those (as required).
I'll post the latest logs below to show the latest error messages:
Quote:
18:06:56:Operating System: Windows (Microsoft Windows 10.0.19045)
18:06:56:CPUs: 11th Gen Intel(R) Core(TM) i5-11400 @ 2.60GHz (Intel)
18:06:56: 1 CPU x 6 cores. 12 logical processors (x64)
18:06:56:GPU: Intel(R) UHD Graphics 730 (1,024 MiB) (Intel Corporation)
18:06:56: Driver: 30.0.101.1273
18:06:56:System RAM: 16 GiB
18:06:56:Target: Windows
18:06:56:BuildConfig: Release
18:06:56:Execution Env: Native
18:06:56:Runtime Env: Production
18:06:56:.NET framework: .NET 7.0.5
18:06:56:App DataDir: C:\ProgramData\CodeProject\AI
18:06:56:Video adapter info:
18:06:57: Intel(R) UHD Graphics 730:
18:06:57: Driver Version 30.0.101.1273
18:06:57: Video Processor Intel(R) UHD Graphics Family
18:06:57:ROOT_PATH = C:\Program Files\CodeProject\AI
18:06:57:RUNTIMES_PATH = C:\Program Files\CodeProject\AI\runtimes
18:06:57:PREINSTALLED_MODULES_PATH = C:\Program Files\CodeProject\AI\preinstalled-modules
18:06:57:MODULES_PATH = C:\Program Files\CodeProject\AI\modules
18:06:57:PYTHON_PATH = \bin\windows\%PYTHON_RUNTIME%\venv\scripts\Python
18:06:57:Data Dir = C:\ProgramData\CodeProject\AI
18:06:57:Server version: 2.1.10-Beta
18:06:57:ModuleRunner Start
18:06:57:Starting Background AI Modules
18:07:00:GetCommandByRuntime: Runtime=python37, Location=Local
18:07:00:Command: C:\Program Files\CodeProject\AI\modules\ObjectDetectionCoral\bin\windows\python37\venv\scripts\Python
18:07:00:
18:07:00:Attempting to start ObjectDetectionCoral with C:\Program Files\CodeProject\AI\modules\ObjectDetectionCoral\bin\windows\python37\venv\scripts\Python "C:\Program Files\CodeProject\AI\modules\ObjectDetectionCoral\objectdetection_coral_adapter.py"
18:07:00:Starting C:\Program Files...ws\python37\venv\scripts\Python "C:\Program Files...ectdetection_coral_adapter.py"
18:07:00:
18:07:00:Module 'ObjectDetection (Coral)' (ID: ObjectDetectionCoral)
18:07:00:Module Path: C:\Program Files\CodeProject\AI\modules\ObjectDetectionCoral
18:07:00:AutoStart: True
18:07:00:Queue: objectdetection_queue
18:07:00:Platforms: windows,linux,linux-arm64,macos,macos-arm64
18:07:00:GPU: Support enabled
18:07:00:Parallelism: 1
18:07:00:Accelerator:
18:07:00:Half Precis.: enable
18:07:00:Runtime: python37
18:07:00:Runtime Loc: Local
18:07:00:FilePath: objectdetection_coral_adapter.py
18:07:00:Pre installed: False
18:07:00:Start pause: 1 sec
18:07:00:LogVerbosity:
18:07:00:Valid: True
18:07:00:Environment Variables
18:07:00:MODELS_DIR = %CURRENT_MODULE_PATH%/assets
18:07:00:MODEL_SIZE = Medium
18:07:00:
18:07:00:Started ObjectDetection (Coral) module
18:07:02:A new version 2.1.11-Beta is available
18:07:02:Current Version is 2.1.10-Beta
18:07:14:objectdetection_coral_adapter.py: E driver/mmio_driver.cc:254] HIB Error. hib_error_status = ffffffffffffffff, hib_first_error_status = ffffffffffffffff
18:07:28:objectdetection_coral_adapter.py: E driver/mmio_driver.cc:254] HIB Error. hib_error_status = ffffffffffffffff, hib_first_error_status = ffffffffffffffff
18:07:28:objectdetection_coral_adapter.py: CPAI_MODULE_REQUIRED_MB not found. Setting to default 0
18:07:28:objectdetection_coral_adapter.py: NUM_THREADS not found. Setting to default 1
18:07:28:objectdetection_coral_adapter.py: MIN_CONFIDENCE not found. Setting to default 0.5
18:07:28:objectdetection_coral_adapter.py: MODULE_PATH: C:\Program Files\CodeProject\AI\modules\ObjectDetectionCoral
18:07:28:objectdetection_coral_adapter.py: MODELS_DIR: C:\Program Files\CodeProject\AI\modules\ObjectDetectionCoral\assets
18:07:28:objectdetection_coral_adapter.py: MODEL_SIZE: medium
18:07:28:objectdetection_coral_adapter.py: CPU_MODEL_NAME: efficientdet_lite3_512_ptq.tflite
18:07:28:objectdetection_coral_adapter.py: TPU_MODEL_NAME: efficientdet_lite3_512_ptq_edgetpu.tflite
18:07:28:objectdetection_coral_adapter.py: Input details: {'name': 'serving_default_images ', 'index': 0, 'shape': array([ 1, 512, 512, 3]), 'shape_signature': array([ 1, 512, 512, 3]), 'dtype': , 'quantization': (0.0078125, 127), 'quantization_parameters': {'scales': array([0.0078125], dtype=float32), 'zero_points': array([127]), 'quantized_dimension': 0}, 'sparsity_parameters': {}}
18:07:28:objectdetection_coral_adapter.py: Output details: {'name': 'StatefulPartitionedCall:31', 'index': 931, 'shape': array([ 1, 25, 4]), 'shape_signature': array([ 1, 25, 4]), 'dtype': , 'quantization': (0.0, 0), 'quantization_parameters': {'scales': array([], dtype=float32), 'zero_points': array([], dtype=int32), 'quantized_dimension': 0}, 'sparsity_parameters': {}}
18:07:28:objectdetection_coral_adapter.py: Timeout connecting to the server
18:07:28:objectdetection_coral_adapter.py: ObjectDetection (Coral) started.ObjectDetection (Coral): ObjectDetection (Coral) started.
18:07:28:ObjectDetection (Coral): ObjectDetection (Coral) started.
18:16:51:Client request 'detect' in queue 'objectdetection_queue' (...fe704d)
18:16:51:Request 'detect' dequeued from 'objectdetection_queue' (...fe704d)
18:16:51:ObjectDetection (Coral): Retrieved objectdetection_queue command
18:16:52:Client request 'detect' in queue 'objectdetection_queue' (...850279)
18:16:52:Request 'detect' dequeued from 'objectdetection_queue' (...850279)
18:16:52:ObjectDetection (Coral): Retrieved objectdetection_queue command
18:16:52:Client request 'detect' in queue 'objectdetection_queue' (...61a35f)
18:16:52:Request 'detect' dequeued from 'objectdetection_queue' (...61a35f)
18:16:52:ObjectDetection (Coral): Retrieved objectdetection_queue command
18:16:53:ObjectDetection (Coral): Queue request for ObjectDetection (Coral) command 'detect' (...850279) took 1019ms
18:16:53:Response received (...850279): The interpreter is in use. Please try again later
18:16:53:ObjectDetection (Coral): Queue request for ObjectDetection (Coral) command 'detect' (...61a35f) took 1025ms
18:16:53:Response received (...61a35f): The interpreter is in use. Please try again later
Logging level
trace
One other question, the Coral database runs locally doesn't it? I currently don't have a dedicated network cable for the BI pc so have to swap it between Main PC and BI PC as required.
Reading the logs (which mean little to me) the main issue seems to be a HIB error status and the interpreter being busy.
One last thought, the log also says Quote: objectdetection_coral_adapter.py: CPAI_MODULE_REQUIRED_MB not found. Setting to default 0
Could it be looking in the wrong place ie PCIE slot by default using CPAI? Mine is in the CPU M.2 slot via an adapter (adapter because the key of the TPU m.2 is different to the motherboard M.2 slot).
modified 22-Aug-23 13:49pm.
|
|
|
|
|
Quote: E driver/mmio_driver.cc:254] HIB Error. hib_error_status = ffffffffffffffff, hib_first_error_status = ffffffffffffffff
This seems to be the issue.
In the Coral module (/modules/ObjectDetectionCoral) should be a edgetpu_runtime folder. In that is the install.bat for the drivers. It would be interesting to re-run that (under admin mode in a Windows Terminal) just to ensure things are good
cheers
Chris Maunder
|
|
|
|
|
Hi Chris,
Didn't know how to do what you said, so I used the Uninstaller in the EdgeRuntime and then reinstalled the Module.
This seems to have it working with some issues:
1. Interpretation tooks 229ms vs 250 for the PC's CPU under Yolo5. This looks plainly wrong as the TPU is both dual and on a direct connection to the CPU's bus via M.2 (non chipset version) so direct path to CPU.
2. A subsequent Interpretation took 1021 ms with no result.
3. After 2. the logs show "The interpreter is in use. Please try again later". This may have been a request from the 2nd camera as I walked through 1 cameras tips to the others and then back again.
Very strange it would timeout though and very long times - remember this is dual TPU so processing should be at least 1/2 what a single TPU would return, no forgoing the direct CPU Bus connection as well circumventing the chipset.
Quote:
18:54:18:Operating System: Windows (Microsoft Windows 10.0.19045)
18:54:18:CPUs: 11th Gen Intel(R) Core(TM) i5-11400 @ 2.60GHz (Intel)
18:54:18: 1 CPU x 6 cores. 12 logical processors (x64)
18:54:18:GPU: Intel(R) UHD Graphics 730 (1,024 MiB) (Intel Corporation)
18:54:18: Driver: 30.0.101.1273
18:54:18:System RAM: 16 GiB
18:54:18:Target: Windows
18:54:18:BuildConfig: Release
18:54:18:Execution Env: Native
18:54:18:Runtime Env: Production
18:54:18:.NET framework: .NET 7.0.5
18:54:18:App DataDir: C:\ProgramData\CodeProject\AI
18:54:18:Video adapter info:
18:54:18: Intel(R) UHD Graphics 730:
18:54:18: Driver Version 30.0.101.1273
18:54:18: Video Processor Intel(R) UHD Graphics Family
18:54:18:ROOT_PATH = C:\Program Files\CodeProject\AI
18:54:18:RUNTIMES_PATH = C:\Program Files\CodeProject\AI\runtimes
18:54:18:PREINSTALLED_MODULES_PATH = C:\Program Files\CodeProject\AI\preinstalled-modules
18:54:18:MODULES_PATH = C:\Program Files\CodeProject\AI\modules
18:54:18:PYTHON_PATH = \bin\windows\%PYTHON_RUNTIME%\venv\scripts\Python
18:54:18:Data Dir = C:\ProgramData\CodeProject\AI
18:54:18:Server version: 2.1.10-Beta
18:54:18:ModuleRunner Start
18:54:18:Starting Background AI Modules
18:54:21:GetCommandByRuntime: Runtime=python37, Location=Local
18:54:21:Command: C:\Program Files\CodeProject\AI\modules\ObjectDetectionCoral\bin\windows\python37\venv\scripts\Python
18:54:21:Starting C:\Program Files...ws\python37\venv\scripts\Python "C:\Program Files...ectdetection_coral_adapter.py"
18:54:21:
18:54:21:Attempting to start ObjectDetectionCoral with C:\Program Files\CodeProject\AI\modules\ObjectDetectionCoral\bin\windows\python37\venv\scripts\Python "C:\Program Files\CodeProject\AI\modules\ObjectDetectionCoral\objectdetection_coral_adapter.py"
18:54:21:
18:54:21:Module 'ObjectDetection (Coral)' (ID: ObjectDetectionCoral)
18:54:21:Module Path: C:\Program Files\CodeProject\AI\modules\ObjectDetectionCoral
18:54:21:AutoStart: True
18:54:21:Queue: objectdetection_queue
18:54:21:Platforms: windows,linux,linux-arm64,macos,macos-arm64
18:54:21:GPU: Support enabled
18:54:21:Parallelism: 1
18:54:21:Accelerator:
18:54:21:Half Precis.: enable
18:54:21:Runtime: python37
18:54:21:Runtime Loc: Local
18:54:21:FilePath: objectdetection_coral_adapter.py
18:54:21:Pre installed: False
18:54:21:Start pause: 1 sec
18:54:21:LogVerbosity:
18:54:21:Valid: True
18:54:21:Environment Variables
18:54:21:MODELS_DIR = %CURRENT_MODULE_PATH%/assets
18:54:21:MODEL_SIZE = Medium
18:54:21:
18:54:21:Started ObjectDetection (Coral) module
18:54:24:objectdetection_coral_adapter.py: CPAI_MODULE_REQUIRED_MB not found. Setting to default 0
18:54:24:objectdetection_coral_adapter.py: NUM_THREADS not found. Setting to default 1
18:54:24:objectdetection_coral_adapter.py: Edge TPU detected
18:54:24:objectdetection_coral_adapter.py: MIN_CONFIDENCE not found. Setting to default 0.5
18:54:24:objectdetection_coral_adapter.py: MODULE_PATH: C:\Program Files\CodeProject\AI\modules\ObjectDetectionCoral
18:54:24:objectdetection_coral_adapter.py: MODELS_DIR: C:\Program Files\CodeProject\AI\modules\ObjectDetectionCoral\assets
18:54:24:objectdetection_coral_adapter.py: MODEL_SIZE: medium
18:54:24:objectdetection_coral_adapter.py: CPU_MODEL_NAME: efficientdet_lite3_512_ptq.tflite
18:54:24:objectdetection_coral_adapter.py: TPU_MODEL_NAME: efficientdet_lite3_512_ptq_edgetpu.tflite
18:54:24:objectdetection_coral_adapter.py: Input details: {'name': 'serving_default_images ', 'index': 0, 'shape': array([ 1, 512, 512, 3]), 'shape_signature': array([ 1, 512, 512, 3]), 'dtype': , 'quantization': (0.0078125, 127), 'quantization_parameters': {'scales': array([0.0078125], dtype=float32), 'zero_points': array([127]), 'quantized_dimension': 0}, 'sparsity_parameters': {}}
18:54:24:objectdetection_coral_adapter.py: Output details: {'name': 'StatefulPartitionedCall:31', 'index': 23, 'shape': array([ 1, 25, 4]), 'shape_signature': array([ 1, 25, 4]), 'dtype': , 'quantization': (0.0, 0), 'quantization_parameters': {'scales': array([], dtype=float32), 'zero_points': array([], dtype=int32), 'quantized_dimension': 0}, 'sparsity_parameters': {}}
18:54:24:Running init for ObjectDetection (Coral)
18:54:24:ObjectDetection (Coral): ObjectDetection (Coral) started.
18:54:24:A new version 2.1.11-Beta is available
18:54:24:Current Version is 2.1.10-Beta
18:57:57:Client request 'detect' in queue 'objectdetection_queue' (...1a68f9)
18:57:57:Request 'detect' dequeued from 'objectdetection_queue' (...1a68f9)
18:57:57:ObjectDetection (Coral): Retrieved objectdetection_queue command
18:57:57:ObjectDetection (Coral): Queue request for ObjectDetection (Coral) command 'detect' (...1a68f9) took 229ms
18:57:57:Response received (...1a68f9): Found person, truck
18:58:03:Client request 'detect' in queue 'objectdetection_queue' (...394efa)
18:58:03:Request 'detect' dequeued from 'objectdetection_queue' (...394efa)
18:58:03:ObjectDetection (Coral): Retrieved objectdetection_queue command
18:58:04:Client request 'detect' in queue 'objectdetection_queue' (...ab17c7)
18:58:04:Request 'detect' dequeued from 'objectdetection_queue' (...ab17c7)
18:58:04:ObjectDetection (Coral): Retrieved objectdetection_queue command
18:58:05:Client request 'detect' in queue 'objectdetection_queue' (...5ad358)
18:58:05:Request 'detect' dequeued from 'objectdetection_queue' (...5ad358)
18:58:05:ObjectDetection (Coral): Retrieved objectdetection_queue command
18:58:05:ObjectDetection (Coral): Queue request for ObjectDetection (Coral) command 'detect' (...ab17c7) took 1021ms
18:58:05:Response received (...ab17c7): The interpreter is in use. Please try again later
|
|
|
|
|
With the dual TPUs you either run inferences in parallel (target one TPU, then target the other, and back again) resulting in higher throughput, but each inference is on a single TPU, or you use pipelining to split a single inference between both TPUs. Unfortunately don't have specific support for multiple TPUs so you won't see a doubling of throughput.
The other thing we've seen is that Coral on Windows is simply slower than any other platform. I don't have an answer as to why, unfortunately.
cheers
Chris Maunder
|
|
|
|
|