It appears the latest version is out. I haven't tried it yet to see if it fixes the issues.
CodeProject.AI Version 2.0
- Thread starter MikeLud1
- Start date
Last edited:
CCTVCam
Known around here
- Sep 25, 2017
- 2,819
- 3,686
Mike do you have a wattage measuring plug and if so, what wattage is your BI PC drawing with a 3060?
I haven't had much success with Coral AI but over the pond here, electricty is really expensive atm. eg. We're paying approx 33 cents per unit + 5% VAT on top making low wattage solution a must have for many.
fenderman
Staff member
- Mar 9, 2014
- 36,892
- 21,407
How many cameras are you running ai on and what resolution? I run CPAI on all my machines using the cpu and/or intel gpu with no issues.
CCTVCam
Known around here
- Sep 25, 2017
- 2,819
- 3,686
Just the two. 2 @ 4K 16800kbs = @ 4,000KBS stream to BI when compressed in camera . The sub streams at 1080P 4,000kbs = 1,000 kbs istreamed to BI.
According to an in the wall power plug, electricty comsumption for the BI PC at idle is around 50w peaking to around 60-70w with activity.
At 60w average that's costing @$170 a year:
Electricity Cost Calculator
Use our simple energy calculator to see the running costs of any electrical items using a range of electricity tariffs
www.sust-it.net
For what it’s worth, I recently rewrote the CPAI Coral implementation and it was accepted into the mainline branch a few days ago. Now it handles multithreaded code, multi TPUs, and segmented inference pipelines. There are a few things left to do, but hopefully this next version is more effective.
fenderman
Staff member
- Mar 9, 2014
- 36,892
- 21,407
What do you mean compressed in camera? I believe you are just confusing bytes and bits...https://ipcamtalk.com/threads/bit-rate.61091/Just the two. 2 @ 4K 16800kbs = @ 4,000KBS stream to BI when compressed in camera . The sub streams at 1080P 4,000kbs = 1,000 kbs istreamed to BI.
According to an in the wall power plug, electricty comsumption for the BI PC at idle is around 50w peaking to around 60-70w with activity.
At 60w average that's costing @$170 a year:
Electricity Cost Calculator
Use our simple energy calculator to see the running costs of any electrical items using a range of electricity tariffs
You can easily run those cams through CPAI on cpu or integrated gpu on any i5 since the second gen
50w is really high for idle..what processor are you using, how many hard drives and what is the cpu load at idle? are you including the cameras power consumption?
Do you have a Graphics Card GPU?Just the two. 2 @ 4K 16800kbs = @ 4,000KBS stream to BI when compressed in camera . The sub streams at 1080P 4,000kbs = 1,000 kbs istreamed to BI.
According to an in the wall power plug, electricty comsumption for the BI PC at idle is around 50w peaking to around 60-70w with activity.
At 60w average that's costing @$170 a year:
Electricity Cost Calculator
Use our simple energy calculator to see the running costs of any electrical items using a range of electricity tariffs
I just installed 2.5RC8.
I have a CORAL dual TPU with an pci adapter that enables both TPUs.
Upon installing the CORAL module and enabling the new multi TPU support I get these errors after a while
I have a CORAL dual TPU with an pci adapter that enables both TPUs.
Upon installing the CORAL module and enabling the new multi TPU support I get these errors after a while
Code:
12:12:07:ObjectDetectionCoral went quietly
12:12:07:
12:12:07:Module 'Object Detection (Coral)' 2.0.3 (ID: ObjectDetectionCoral)
12:12:07:Valid: True
12:12:07:Module Path: <root>\modules\ObjectDetectionCoral
12:12:07:AutoStart: True
12:12:07:Queue: objectdetection_queue
12:12:07:Runtime: python3.9
12:12:07:Runtime Loc: Local
12:12:07:FilePath: objectdetection_coral_adapter.py
12:12:07:Pre installed: False
12:12:07:Start pause: 1 sec
12:12:07:LogVerbosity:
12:12:07:Platforms: all
12:12:07:GPU Libraries: installed if available
12:12:07:GPU Enabled: enabled
12:12:07:Parallelism: 1
12:12:07:Accelerator:
12:12:07:Half Precis.: enable
12:12:07:Environment Variables
12:12:07:CPAI_CORAL_MULTI_TPU = True
12:12:07:CPAI_CORAL_USE_YOLO = true
12:12:07:MODELS_DIR = <root>\modules\ObjectDetectionCoral\assets
12:12:07:MODEL_SIZE = medium
12:12:07:
12:12:07:Started Object Detection (Coral) module
12:14:09:objectdetection_coral_adapter.py: WARNING:root:No work in 120.0 seconds, watchdog shutting down TPU runners!
12:14:09:objectdetection_coral_adapter.py: WARNING: Logging before InitGoogleLogging() is written to STDERR
12:14:09:objectdetection_coral_adapter.py: I20240119 12:14:09.623443 15248 pipelined_model_runner.cc:171] Thread: 15248 receives empty request
12:14:09:objectdetection_coral_adapter.py: I20240119 12:14:09.623443 15248 pipelined_model_runner.cc:244] Thread: 15248 is shutting down the pipeline...
12:14:09:objectdetection_coral_adapter.py: I20240119 12:14:09.623443 15248 pipelined_model_runner.cc:254] Thread: 15248 Pipeline is off.Looks like no work is reaching the module, so there is some connection upstream of that that’s missing. I’d be curious if you can turn on debug level logging, but I don’t know if there’s an option for it
OK, I set the log level to TRACE, stopped yolo and started coral
Server page shows "lost contact"
Did a refresh (CTRL F5), this is the log
started coral again...
Server shows "Lost Contact"
redoing the coral settings to "doubletpu"
it seems there were some requests from BI...
again, these errors
Server page shows "lost contact"
Did a refresh (CTRL F5), this is the log
Code:
16:41:58:Update ObjectDetectionYOLOv5Net. Setting AutoStart=false
16:41:58:Sending shutdown request to ObjectDetectionYOLOv5Net/ObjectDetectionYOLOv5Net
16:41:58:Client request 'Quit' in queue 'objectdetection_queue' (...08ff56)
16:41:58:Request 'Quit' dequeued from 'objectdetection_queue' (...08ff56)
16:41:58:ObjectDetectionYOLOv5Net.exe: Application is shutting down...
16:41:58:ObjectDetectionYOLOv5Net.exe: Unable to get request from objectdetection_queue for ObjectDetectionYOLOv5Net
16:41:58:ObjectDetectionYOLOv5Net.exe: Unable to get request from objectdetection_queue for ObjectDetectionYOLOv5Net
16:41:58:ObjectDetectionYOLOv5Net.exe: Unable to get request from objectdetection_queue for ObjectDetectionYOLOv5Net
16:41:58:ObjectDetectionYOLOv5Net.exe: Unable to get request from objectdetection_queue for ObjectDetectionYOLOv5Net
16:41:58:ObjectDetectionYOLOv5Net.exe: Unable to get request from objectdetection_queue for ObjectDetectionYOLOv5Net
16:41:58:ObjectDetectionYOLOv5Net.exe: Unable to get request from objectdetection_queue for ObjectDetectionYOLOv5Net
16:41:58:ObjectDetectionYOLOv5Net.exe: Unable to get request from objectdetection_queue for ObjectDetectionYOLOv5Net
16:41:58:ObjectDetectionYOLOv5Net.exe: Unable to get request from objectdetection_queue for ObjectDetectionYOLOv5Net
16:41:58:ObjectDetectionYOLOv5Net.exe: Unable to get request from objectdetection_queue for ObjectDetectionYOLOv5Net
16:41:58:ObjectDetectionYOLOv5Net.exe: Unable to get request from objectdetection_queue for ObjectDetectionYOLOv5Net
16:41:58:ObjectDetectionYOLOv5Net.exe: Unable to get request from objectdetection_queue for ObjectDetectionYOLOv5Net
16:41:58:ObjectDetectionYOLOv5Net.exe: Shutdown signal received. Ending loop
16:41:58:ObjectDetectionYOLOv5Net.exe: Shutdown signal received. Ending loop
16:41:58:ObjectDetectionYOLOv5Net.exe: Shutdown signal received. Ending loop
16:41:58:ObjectDetectionYOLOv5Net.exe: Shutdown signal received. Ending loop
16:41:58:ObjectDetectionYOLOv5Net.exe: Shutdown signal received. Ending loop
16:41:59:Module ObjectDetectionYOLOv5Net has shutdown
16:41:59:ObjectDetectionYOLOv5Net.exe: has exited
16:42:31:ObjectDetectionYOLOv5Net went quietly
16:42:41:Running module using: C:\Program Files\CodeProject\AI\modules\ObjectDetectionCoral\bin\windows\python39\venv\Scripts\python
16:42:41:Update ObjectDetectionCoral. Setting AutoStart=true
16:42:41:Starting C:\Program Files...ws\python39\venv\Scripts\python "C:\Program Files...ectdetection_coral_adapter.py"
16:42:41:
16:42:41:
16:42:41:Attempting to start ObjectDetectionCoral with C:\Program Files\CodeProject\AI\modules\ObjectDetectionCoral\bin\windows\python39\venv\Scripts\python "C:\Program Files\CodeProject\AI\modules\ObjectDetectionCoral\objectdetection_coral_adapter.py"
16:42:41:Module 'Object Detection (Coral)' 2.0.3 (ID: ObjectDetectionCoral)
16:42:41:Valid: True
16:42:41:Module Path: <root>\modules\ObjectDetectionCoral
16:42:41:AutoStart: True
16:42:41:Queue: objectdetection_queue
16:42:41:Runtime: python3.9
16:42:41:Runtime Loc: Local
16:42:41:FilePath: objectdetection_coral_adapter.py
16:42:41:Pre installed: False
16:42:41:Start pause: 1 sec
16:42:41:LogVerbosity:
16:42:41:Platforms: all
16:42:41:GPU Libraries: installed if available
16:42:41:GPU Enabled: enabled
16:42:41:Parallelism: 1
16:42:41:Accelerator:
16:42:41:Half Precis.: enable
16:42:41:Environment Variables
16:42:41:CPAI_CORAL_MULTI_TPU = True
16:42:41:CPAI_CORAL_USE_YOLO = true
16:42:41:MODELS_DIR = <root>\modules\ObjectDetectionCoral\assets
16:42:41:MODEL_SIZE = medium
16:42:41:
16:42:41:Started Object Detection (Coral) module
16:44:44:objectdetection_coral_adapter.py: WARNING:root:No work in 120.0 seconds, watchdog shutting down TPU runners!
16:44:44:objectdetection_coral_adapter.py: WARNING: Logging before InitGoogleLogging() is written to STDERR
16:44:44:objectdetection_coral_adapter.py: I20240119 16:44:44.554811 11144 pipelined_model_runner.cc:171] Thread: 11144 receives empty request
16:44:44:objectdetection_coral_adapter.py: I20240119 16:44:44.554811 11144 pipelined_model_runner.cc:244] Thread: 11144 is shutting down the pipeline...
16:44:44:objectdetection_coral_adapter.py: I20240119 16:44:44.554811 11144 pipelined_model_runner.cc:254] Thread: 11144 Pipeline is off.started coral again...
Code:
16:47:01:Forcing shutdown of python/ObjectDetectionCoral
16:47:01:Module ObjectDetectionCoral has shutdown
16:47:01:Waiting for ObjectDetectionCoral to end.
16:47:01:objectdetection_coral_adapter.py: has exited
16:47:01:ObjectDetectionCoral ended after 111 ms
16:47:01:Running module using: C:\Program Files\CodeProject\AI\modules\ObjectDetectionCoral\bin\windows\python39\venv\Scripts\python
16:47:01:Starting C:\Program Files...ws\python39\venv\Scripts\python "C:\Program Files...ectdetection_coral_adapter.py"
16:47:01:
16:47:01:
16:47:01:Attempting to start ObjectDetectionCoral with C:\Program Files\CodeProject\AI\modules\ObjectDetectionCoral\bin\windows\python39\venv\Scripts\python "C:\Program Files\CodeProject\AI\modules\ObjectDetectionCoral\objectdetection_coral_adapter.py"
16:47:01:Module 'Object Detection (Coral)' 2.0.3 (ID: ObjectDetectionCoral)
16:47:01:Valid: True
16:47:01:Module Path: <root>\modules\ObjectDetectionCoral
16:47:01:AutoStart: True
16:47:01:Queue: objectdetection_queue
16:47:01:Runtime: python3.9
16:47:01:Runtime Loc: Local
16:47:01:FilePath: objectdetection_coral_adapter.py
16:47:01:Pre installed: False
16:47:01:Start pause: 1 sec
16:47:01:LogVerbosity:
16:47:01:Platforms: all
16:47:01:GPU Libraries: installed if available
16:47:01:GPU Enabled: enabled
16:47:01:Parallelism: 1
16:47:01:Accelerator:
16:47:01:Half Precis.: enable
16:47:01:Environment Variables
16:47:01:CPAI_CORAL_MULTI_TPU = True
16:47:01:CPAI_CORAL_USE_YOLO = true
16:47:01:MODELS_DIR = <root>\modules\ObjectDetectionCoral\assets
16:47:01:MODEL_SIZE = medium
16:47:01:
16:47:01:Started Object Detection (Coral) moduleServer shows "Lost Contact"
redoing the coral settings to "doubletpu"
Code:
16:48:54:Client request 'Quit' in queue 'objectdetection_queue' (...9ad588)
16:48:54:Update ObjectDetectionCoral. Setting CPAI_CORAL_MULTI_TPU=True
16:48:54:Sending shutdown request to python/ObjectDetectionCoral
16:49:03:objectdetection_coral_adapter.py: WARNING:root:No work in 120.0 seconds, watchdog shutting down TPU runners!
16:49:03:objectdetection_coral_adapter.py: WARNING: Logging before InitGoogleLogging() is written to STDERR
16:49:03:objectdetection_coral_adapter.py: I20240119 16:49:03.168071 14056 pipelined_model_runner.cc:171] Thread: 14056 receives empty request
16:49:03:objectdetection_coral_adapter.py: I20240119 16:49:03.168071 14056 pipelined_model_runner.cc:244] Thread: 14056 is shutting down the pipeline...
16:49:03:objectdetection_coral_adapter.py: I20240119 16:49:03.168071 14056 pipelined_model_runner.cc:254] Thread: 14056 Pipeline is off.
16:49:27:Forcing shutdown of python/ObjectDetectionCoral
16:49:27:Module ObjectDetectionCoral has shutdown
16:49:27:Waiting for ObjectDetectionCoral to end.
16:49:27:Running module using: C:\Program Files\CodeProject\AI\modules\ObjectDetectionCoral\bin\windows\python39\venv\Scripts\python
16:49:27:objectdetection_coral_adapter.py: has exited
16:49:27:ObjectDetectionCoral ended after 106 ms
16:49:27:Starting C:\Program Files...ws\python39\venv\Scripts\python "C:\Program Files...ectdetection_coral_adapter.py"
16:49:27:
16:49:27:Attempting to start ObjectDetectionCoral with C:\Program Files\CodeProject\AI\modules\ObjectDetectionCoral\bin\windows\python39\venv\Scripts\python "C:\Program Files\CodeProject\AI\modules\ObjectDetectionCoral\objectdetection_coral_adapter.py"
16:49:27:
16:49:27:Module 'Object Detection (Coral)' 2.0.3 (ID: ObjectDetectionCoral)
16:49:27:Valid: True
16:49:27:Module Path: <root>\modules\ObjectDetectionCoral
16:49:27:AutoStart: True
16:49:27:Queue: objectdetection_queue
16:49:27:Runtime: python3.9
16:49:27:Runtime Loc: Local
16:49:27:FilePath: objectdetection_coral_adapter.py
16:49:27:Pre installed: False
16:49:27:Start pause: 1 sec
16:49:27:LogVerbosity:
16:49:27:Platforms: all
16:49:27:GPU Libraries: installed if available
16:49:27:GPU Enabled: enabled
16:49:27:Parallelism: 1
16:49:27:Accelerator:
16:49:27:Half Precis.: enable
16:49:27:Environment Variables
16:49:27:CPAI_CORAL_MULTI_TPU = True
16:49:27:CPAI_CORAL_USE_YOLO = true
16:49:27:MODELS_DIR = <root>\modules\ObjectDetectionCoral\assets
16:49:27:MODEL_SIZE = medium
16:49:27:
16:49:27:Started Object Detection (Coral) module
16:50:04:Client request 'detect' in queue 'objectdetection_queue' (...a260f6)
16:50:04:Client request 'detect' in queue 'objectdetection_queue' (...d5ea6c)
16:50:05:Client request 'detect' in queue 'objectdetection_queue' (...586cf6)
16:50:06:Client request 'detect' in queue 'objectdetection_queue' (...b7e42b)it seems there were some requests from BI...
Code:
16:51:29:objectdetection_coral_adapter.py: WARNING:root:No work in 120.0 seconds, watchdog shutting down TPU runners!
16:51:29:objectdetection_coral_adapter.py: WARNING: Logging before InitGoogleLogging() is written to STDERR
16:51:29:objectdetection_coral_adapter.py: I20240119 16:51:29.429220 1400 pipelined_model_runner.cc:171] Thread: 1400 receives empty request
16:51:29:objectdetection_coral_adapter.py: I20240119 16:51:29.429220 1400 pipelined_model_runner.cc:244] Thread: 1400 is shutting down the pipeline...
16:51:29:objectdetection_coral_adapter.py: I20240119 16:51:29.429977 1400 pipelined_model_runner.cc:254] Thread: 1400 Pipeline is off.disabling multi tpu support leads to the following log. Alas, nothing detected when I try test pics on the explorer
Please let me know if I should test anything else.
Thank you for working on this!
Please let me know if I should test anything else.
Thank you for working on this!
Code:
16:53:05:Client request 'Quit' in queue 'objectdetection_queue' (...700ea6)
16:53:05:Update ObjectDetectionCoral. Setting CPAI_CORAL_MULTI_TPU=False
16:53:05:Sending shutdown request to python/ObjectDetectionCoral
16:53:38:Forcing shutdown of python/ObjectDetectionCoral
16:53:38:Module ObjectDetectionCoral has shutdown
16:53:38:Waiting for ObjectDetectionCoral to end.
16:53:38:Running module using: C:\Program Files\CodeProject\AI\modules\ObjectDetectionCoral\bin\windows\python39\venv\Scripts\python
16:53:38:objectdetection_coral_adapter.py: has exited
16:53:38:ObjectDetectionCoral ended after 109 ms
16:53:38:Starting C:\Program Files...ws\python39\venv\Scripts\python "C:\Program Files...ectdetection_coral_adapter.py"
16:53:38:
16:53:38:Module 'Object Detection (Coral)' 2.0.3 (ID: ObjectDetectionCoral)
16:53:38:Valid: True
16:53:38:Module Path: <root>\modules\ObjectDetectionCoral
16:53:38:
16:53:38:Attempting to start ObjectDetectionCoral with C:\Program Files\CodeProject\AI\modules\ObjectDetectionCoral\bin\windows\python39\venv\Scripts\python "C:\Program Files\CodeProject\AI\modules\ObjectDetectionCoral\objectdetection_coral_adapter.py"
16:53:38:AutoStart: True
16:53:38:Queue: objectdetection_queue
16:53:38:Runtime: python3.9
16:53:38:Runtime Loc: Local
16:53:38:FilePath: objectdetection_coral_adapter.py
16:53:38:Pre installed: False
16:53:38:Start pause: 1 sec
16:53:38:LogVerbosity:
16:53:38:Platforms: all
16:53:38:GPU Libraries: installed if available
16:53:38:GPU Enabled: enabled
16:53:38:Parallelism: 1
16:53:38:Accelerator:
16:53:38:Half Precis.: enable
16:53:38:Environment Variables
16:53:38:CPAI_CORAL_MULTI_TPU = False
16:53:38:CPAI_CORAL_USE_YOLO = true
16:53:38:MODELS_DIR = <root>\modules\ObjectDetectionCoral\assets
16:53:38:MODEL_SIZE = medium
16:53:38:
16:53:38:Started Object Detection (Coral) module
16:53:40:objectdetection_coral_adapter.py: MODULE_PATH: C:\Program Files\CodeProject\AI\modules\ObjectDetectionCoral
16:53:40:objectdetection_coral_adapter.py: Running init for Object Detection (Coral)
16:53:40:objectdetection_coral_adapter.py: MODELS_DIR: C:\Program Files\CodeProject\AI\modules\ObjectDetectionCoral\assets
16:53:40:objectdetection_coral_adapter.py: MODEL_SIZE: medium
16:53:40:objectdetection_coral_adapter.py: CPU_MODEL_NAME: efficientdet_lite2_448_ptq.tflite
16:53:40:objectdetection_coral_adapter.py: TPU_MODEL_NAME: efficientdet_lite2_448_ptq_edgetpu.tflite
16:53:40:objectdetection_coral_adapter.py: Edge TPU detected
16:53:40:objectdetection_coral_adapter.py: Using Edge TPU
16:53:40:objectdetection_coral_adapter.py: Input details: {'name': 'serving_default_images:0', 'index': 0, 'shape': array([ 1, 448, 448, 3]), 'shape_signature': array([ 1, 448, 448, 3]), 'dtype': , 'quantization': (0.0078125, 127), 'quantization_parameters': {'scales': array([0.0078125], dtype=float32), 'zero_points': array([127]), 'quantized_dimension': 0}, 'sparsity_parameters': {}}
16:53:40:objectdetection_coral_adapter.py: Output details: {'name': 'StatefulPartitionedCall:3', 'index': 16, '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': {}}
16:53:40:Request 'Quit' dequeued from 'objectdetection_queue' (...eaeb2f)
16:53:40:Request 'Quit' dequeued from 'objectdetection_queue' (...9ad588)
16:53:40:Request 'detect' dequeued from 'objectdetection_queue' (...a260f6)
16:53:40:Request 'detect' dequeued from 'objectdetection_queue' (...d5ea6c)
16:53:40:Request 'detect' dequeued from 'objectdetection_queue' (...586cf6)
16:53:40:Request 'detect' dequeued from 'objectdetection_queue' (...b7e42b)
16:53:40:Request 'detect' dequeued from 'objectdetection_queue' (...402547)
16:53:40:Request 'detect' dequeued from 'objectdetection_queue' (...0cad90)
16:53:40:Request 'detect' dequeued from 'objectdetection_queue' (...500211)
16:53:40:Request 'detect' dequeued from 'objectdetection_queue' (...da423a)
16:53:40:Request 'detect' dequeued from 'objectdetection_queue' (...6d4fb9)
16:53:40:Request 'detect' dequeued from 'objectdetection_queue' (...ad10d7)
16:53:40:Request 'detect' dequeued from 'objectdetection_queue' (...79541c)
16:53:40:Request 'detect' dequeued from 'objectdetection_queue' (...8f98f1)
16:53:40:Object Detection (Coral): Object Detection (Coral) started.
16:53:40:Request 'Quit' dequeued from 'objectdetection_queue' (...700ea6)When multiTPU is set to false is should follow the previous Coral TPU code path and touch very little changed code. Is that working at all for you? I see nothing in your logs showing that it's actually attempting to make an inference.
CCTVCam
Known around here
- Sep 25, 2017
- 2,819
- 3,686
Spec is I5-11400, 16GB DDR4, Single 4TB WD Surveillance Drive, CPU Cooler, 2 fans with PWM that sit mostly idle. Coral Dual TPU fitted in M.2 (5w) but not currently in use.
This was he build thread in case I missed anything out:
Blue Iris PC Build
As promised this is my PC Build Thread. I'll divide into sections so if you want to you can skip parts that don't interest you (hopefully it all does!). It's not intended as a definitive "this is the pc you need / must build". It represents my choices based on what I perceived to be the best...
Maybe it's just Intel 11 gen as later processors are supposed to be power hungry, or maybe is something a miss:
Here's a screenshot, No UI just BI running in the background, no AV or other Apps running - 45.5 watts:
No. Integrated graphics + Coral Dual TPU.
FWIW, the Dual TPU shouldn't use 5w unless its going all out. Do you have a heatsink on it? I ended up going through a number of different iterations before finding something that would keep it under 85C when under load. Above that temperature, it throttles itself.
coral.ai
See page 5 for power usage, thermal management starts on page 7.
You can monitor TPU temperatures here:
coral.ai
M.2 Accelerator with Dual Edge TPU datasheet | Coral
Technical details for the Coral M.2 Accelerator with Dual Edge TPU
coral.ai
You can monitor TPU temperatures here:
Manage the PCIe module temperature | Coral
Learn how to read the temperature from the Edge TPU and configure frequency scaling when using the Edge TPU over PCIe interface.
coral.ai
Last edited:
Fun fact: with half of my cores disabled (3/6) on my i5-8500 system and five TPUs, I run YOLOv5l (45.5 M Params) inferences at the rate of one every 35.9 ms with the new multi-TPU code. However, each of the 16 threads spends on average ~455 ms waiting for inferences to happen. With YOLOv5n (1.9 M Params), these numbers change to a throughput of one every 32.72 ms and a wait time of ~90 ms. Why does the wait time drop, but the throughput stays roughly the same? I don't know.
Hey.
Anyone running 4060 series Gpu? Testing a 4060 (newest game drivers,12.3 cuda) the detection times seems a bit slow. Running YOLOv5 6.2.. 75-105ms
.NET using ML is a bit faster in around 55ms.
I’ve used 3060-3070 in the past. Super fast times. I do know the CUDA count is better on the 30 series lower end gpu. Just curious what other members are getting.
Anyone running 4060 series Gpu? Testing a 4060 (newest game drivers,12.3 cuda) the detection times seems a bit slow. Running YOLOv5 6.2.. 75-105ms
.NET using ML is a bit faster in around 55ms.
I’ve used 3060-3070 in the past. Super fast times. I do know the CUDA count is better on the 30 series lower end gpu. Just curious what other members are getting.