Intel® Distribution of OpenVINO™ Toolkit
Community assistance about the Intel® Distribution of OpenVINO™ toolkit, OpenCV, and all aspects of computer vision-related on Intel® platforms.
6480 Discussions

Intel Profiling Tool and Power measurement

Shravanthi
Beginner
877 Views

Hello,

 

We have collected inferences(throughput and latency) using OpenVINO toolkit for some of the models but along with this we wanted to meassure power on the device while running these models and also check which layers are falling back on CPU during execution. For power meassurement we tried installing PCM, but we are facing issues with installations and for profiling we have successfully installed VTune Profiler but we have to pass executable (.exe) to capture performance analysis. Could you please help if there is any specific tool for power meassurement and profiling.

 

Regards,

Shravanthi J 

0 Kudos
1 Solution
Megat_Intel
Moderator
848 Views

Hi Shravanthi J,

Thank you for reaching out to us.

 

As of now, OpenVINO™ Toolkit does not yet include features to measure power consumption during inference. You might need to rely on using external equipment to perform the required measurements.

 

For your information, you can use the code below to figure out which layers are falling back to the CPU device:

from openvino.runtime import Core

core = Core()
model = core.read_model("<model_path>")

device_name = "HETERO:GPU,CPU"
compiled_model = core.compile_model(model, device_name)

supported_ops = core.query_model(model, device_name)

for layer in supported_ops:
   print ( " {}: {}" .format(layer, supported_ops[layer]))

fallback .png

 

On the other hand, for profiling, OpenVINO™ has the capabilities for Performance Analysis of key stages such as read time and load time. Most of the modules and features have been tagged with Intel ITT counters, which allows us to measure the performance of these components. The analysis uses the Intel SEAPI and the Intel Vtune Profiler. You can refer to the Performance analysis using ITT counters page for more information.

 

 

Regards,

Megat

 

View solution in original post

0 Kudos
2 Replies
Megat_Intel
Moderator
849 Views

Hi Shravanthi J,

Thank you for reaching out to us.

 

As of now, OpenVINO™ Toolkit does not yet include features to measure power consumption during inference. You might need to rely on using external equipment to perform the required measurements.

 

For your information, you can use the code below to figure out which layers are falling back to the CPU device:

from openvino.runtime import Core

core = Core()
model = core.read_model("<model_path>")

device_name = "HETERO:GPU,CPU"
compiled_model = core.compile_model(model, device_name)

supported_ops = core.query_model(model, device_name)

for layer in supported_ops:
   print ( " {}: {}" .format(layer, supported_ops[layer]))

fallback .png

 

On the other hand, for profiling, OpenVINO™ has the capabilities for Performance Analysis of key stages such as read time and load time. Most of the modules and features have been tagged with Intel ITT counters, which allows us to measure the performance of these components. The analysis uses the Intel SEAPI and the Intel Vtune Profiler. You can refer to the Performance analysis using ITT counters page for more information.

 

 

Regards,

Megat

 

0 Kudos
Megat_Intel
Moderator
744 Views

Hi Shravanthi J,

Thank you for your question. This thread will no longer be monitored since this issue has been resolved. If you need additional information from Intel, please submit a new question.

 

 

Regards,

Megat

 

0 Kudos
Reply