Performance Calculation failed for Syntiant model

Report bugs
, ,
#1

Question/Issue:
Performance Calculation failed for Syntiant model

Project ID:
m / box2mmHole

Context/Use case:
running 10min simulated data

Summary:
Using 8 threads for inference.
INFO: Created TensorFlow Lite XNNPACK delegate for CPU.
Processing 1743 windows…
2024-06-23 07:41:49.119079: E tensorflow/compiler/xla/stream_executor/cuda/cuda_driver.cc:267] failed call to cuInit: UNKNOWN ERROR (34)
Progress: 64%
DSP took 0:00:03, inference took 0:00:01
Complete

followed by Error could not save config

Steps to Reproduce:
Use stored project “PerformanceCalibration Error”
Run Performance Calibrarion

Expected Results:
display graph of results

Actual Results:
No graph and could not save
and get error trying to save config, but refresh page suggests config was saved
as box now shows “clear save config”

Reproducibility:

  • [ ] Always

Environment:

  • Nicla Voice

  • Webbase edge impulse

  • stored project “PerformanceCalibration Error”

  • Increased first 2D conv to 16 filters

#2

Hi @drmpf,

Please note that Performance Calibration does not work for the Syntiant models, as Syntiant has their own post-processing methods. Please see here for more information about how to use the Syntiant post-processing calibration: Responding to your voice - Syntiant - RC Commands | Edge Impulse Documentation

#3

Thanks for that.
Performance Calibration used to work last week with Syntiant settings and then just stopped working this week with that error msg.