Unable to generate features

When I run the spectral feature generation on data on edge impulse I get the following error

Creating windows from 27 files...
[ 1/27] Creating windows from files...
[27/27] Creating windows from files...
Created 81 windows: M: 42, O: 39

Scheduling job in cluster...
Job started
Creating features
[ 1/81] Creating features...
[81/81] Creating features...
Created features

Attached to job 188264...

No data received from server for 30 seconds...

Job failed (see above)

Kindly advise

Interesting, that looks like a finished job to me. It seems that the job finished faster than the log attachment came through. You can see the feature explorer right? If that’s the case then the job ran successful anyway.

I cant see the any features or graph in the feature explorer

No features generated yet.

What’s your project ID? I’ll take a look.

Hi When I run again I get the following error

Creating job... 

Failed to create job: Cannot start job. There already is a job with key "dsp-2" running.

Job failed (see above)

My Project is vibDemoFieldDevice

Creating job... OK (ID: 188272)

Not generating new features: features already generated and no options or files have changed.

Job completed

So this is what I see. I think this might be related to the distance to the data center, if the job finishes sometimes the last job messages don’t come through…

@janjongboom
Thank you for logging into my project for support. I see that when I increase the window size from 2800ms that you set to 4000ms, It now shows that necessary frames are not there. Can you share how the calculation of window size is done so that Spectral Features can be generated?

@paulphilip your items in the dataset are 3 seconds long, so your window should be less than that; otherwise you’ll get these errors indeed.

@janjongboom Please excuse that I got the project wrong. It is TESTDATAUPLOAD and the data length is 5s. The errors mentioned above is for the project above.

Interesting. I’ve killed the job for you now, and will make something where you won’t get stuck.

@mathijs could you look at the logs to see why this job was not marked as finished? Job ID 188264.

@janjongboom @mathijs
Hi I was trying to build the model for project 5392 but it is not getting scheduled for build

Still waiting for job to be scheduled...
Job did not start within 60 seconds

Received cancellation, stopping job in cluster...

Attached to job 255795...
No data received from server for 30 seconds...

Job failed (see above)

When I rerun again I get the following error.


Failed to create job: Cannot start job. There already is a job with key "dsp-3" running.

Job failed (see above)

Kindly advise on how to avoid this problem.
/cc @Hardik

@janjongboom @mathijs

Hi I used the cancel job api to cancel the job and reschedule it once again. I am unable to do it. Kindly advise.

Creating job... OK (ID: 255797)

Retraining Spectral features...
Still waiting for job to be scheduled...
Still waiting for job to be scheduled...
Still waiting for job to be scheduled...
Still waiting for job to be scheduled...
Still waiting for job to be scheduled...
Still waiting for job to be scheduled...
Still waiting for job to be scheduled...
Still waiting for job to be scheduled...
Still waiting for job to be scheduled...
Still waiting for job to be scheduled...
Still waiting for job to be scheduled...
Still waiting for job to be scheduled...
Still waiting for job to be scheduled...
Still waiting for job to be scheduled...
Still waiting for job to be scheduled...
Still waiting for job to be scheduled...
Still waiting for job to be scheduled...
Still waiting for job to be scheduled...
Still waiting for job to be scheduled...
Job did not start within 60 seconds

No data received from server for 60 seconds...

Job failed (see above)

/cc @Hardik

Hi @paulphilip, we had an outage this morning (events to/from job containers were not routed), this has now been resolved!

Hi @janjongboom

I was able to build the model. Now I am facing a new error when I am trying to run the model to classify on a test data.

FreqEINotMatching

Please advise.

/cc @Hardik

Hi, we rolled out a change to prevent samples with a different frequency to be classified (to avoid mixing frequencies), but apparently there is some rounding differences involved when calculating the frequency. Will push out a fix today.

@paulphilip @Hardik this has now been rolled out!